CANCER BIOMARKERS

Abstract

The present invention relates to a method of screening for a cancer selected from the group consisting of prostate cancer, thyroid cancer, colon cancer, rectum cancer, lung cancer, uterine cancer, breast cancer, pancreatic cancer, bladder cancer, liver cancer, bile duct cancer, stomach cancer, oesophageal cancer, head and neck cancer, brain cancer, blood cancer, ovarian cancer, skin cancer, melanoma and a neuroendocrine tumour in a subject, said method comprising determining the level and/or chemical composition of one or more of the glycosaminoglycans (GAGs) chondroitin sulfate (CS), heparan sulfate (HS), and hyaluronic acid (HA) in a body fluid sample, wherein said sample has been obtained from said subject.

Claims

1. A method of screening for a cancer selected from the group consisting of prostate cancer, thyroid cancer, colon cancer, rectum cancer, lung cancer, uterine cancer, breast cancer, pancreatic cancer, bladder cancer, liver cancer, bile duct cancer, stomach cancer, oesophageal cancer, head and neck cancer, brain cancer, blood cancer, ovarian cancer, skin cancer, melanoma and a neuroendocrine tumour in a subject, said method comprising determining the level and/or chemical composition of one or more of the glycosaminoglycans (GAGs) chondroitin sulfate (CS), heparan sulfate (HS), and hyaluronic acid (HA) in a body fluid sample, wherein said sample has been obtained from said subject.

2. The method of claim 1, wherein an altered level and/or chemical composition of chondroitin sulfate (CS) and/or heparan sulfate (HS) and/or hyaluronic acid (HA) in said sample in comparison to a control level and/or chemical composition is indicative of said cancer in said subject.

3. The method of claim 1 or claim 2, wherein said determination of the chemical composition comprises determining the level in the sample of one or more GAG properties selected from the group consisting of: one or more of the specific sulfated or unsulfated forms of CS or HS disaccharides, the fraction of sulfated disaccharides of HS out of the total HS disaccharides present (charge HS), the fraction of sulfated disaccharides of CS out of the total CS disaccharides present (charge CS), the total concentration of CS, the total concentration of HS and the total concentration of HA.

4. The method of any one of claims 1 to 3, wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties selected from the group consisting of: the specific sulfated or unsulfated forms of CS or HS disaccharides selected from the group consisting of: 0s CS, 2s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, 0s HS, 2s HS, 6s HS, 2s6s HS, Ns HS, Ns2s HS, Ns6s HS, Tris HS, the relative level of 4s CS with respect to 6s CS, the relative level of 6s CS with respect to 0s CS and the relative level of 4s CS with respect to 0s CS; the fraction of sulfated disaccharides of HS out of the total HS disaccharides present (charge HS); the fraction of sulfated disaccharides of CS out of the total CS disaccharides present (charge CS); the total concentration of CS (CS tot); the total concentration of HS (HS tot); and the total concentration of HA (HA tot).

5. The method of claim 4, wherein said method comprises determining the level of up to 8, or all 8, of the sulfated and unsulfated CS forms: 0s CS, 2s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS and Tris CS, optionally together with total CS and/or charge CS; and/or wherein said method comprises determining the level of up to 8, or all 8, of the sulfated and unsulfated HS forms: 0s HS, 2s HS, 6s HS, 2s6s HS, Ns HS, Ns2s HS, Ns6s HS and Tris HS, optionally together with total HS and/or charge HS; and/or wherein said method comprises determining the total concentration of HA.

6. The method of any one of claims 1 to 5, wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties selected from the group consisting of: 6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/6s CS, 0s CS, Ns2s HS, 4s CS.

7. The method of any one of claims 1 to 6, wherein said body fluid is a blood sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties 6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/6s CS and 0s CS.

8. The method of any one of claims 1 to 6, wherein said body fluid is a urine sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or both of the GAG properties Ns2s Hs and 4s CS.

9. The method of any one claims 1 to 8, wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, the ratio 4s CS/6s CS, Charge CS, Ns HS, 4s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, 0s HS, Charge HS.

10. The method of any one of claim 1 to 7 or 9, wherein said body fluid is a blood sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, the ratio 4s CS/6s CS, Charge CS, Ns HS.

11. The method of any one of claim 1 to 6, 8 or 9 wherein said body fluid is a urine sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties 4s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, 0s HS, Charge HS.

12. The method of any one of claim 1 to 7, 9 or 10, wherein said body fluid sample is a blood sample and wherein an increase in the level of one or more or all of the GAG properties selected from the group consisting of 6s CS and Charge CS, in comparison to a control level, is indicative of cancer in said subject.

13. The method of any one of claim 1 to 7, 9, 10 or 12, wherein said body fluid sample is a blood sample and wherein a decrease in the level of one or more or all of the GAG properties selected from the group consisting of 0s CS, the ratio 4s CS/6s CS or Ns HS, in comparison to a control level, is indicative of cancer in said subject.

14. The method of any one of claim 1 to 6, 8, 9 or 11, wherein said body fluid sample is a urine sample and wherein an increase in the level of Charge HS, in comparison to a control level, is indicative of cancer in said subject.

15. The method of any one of claim 1 to 6, 8, 9, 11 or 14, wherein said body fluid sample is a urine sample and wherein a decrease in the level of one or more or all of the GAG properties selected from the group consisting of 4s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS or 0s HS, is indicative of cancer in said subject.

16. The method of any one of claims 1 to 5, wherein said body fluid sample is blood and wherein said method comprises determining the level in the sample of one or more of the GAG properties Charge CS, Total CS, 6s CS, 4s CS, 2s HS, 0s HS.

17. The method of any one of claims 1 to 5, wherein said body fluid sample is blood and wherein said method comprises determining the level in the sample of one or more of the GAG properties Charge CS, Total CS, Total HS, 0s CS, 2s CS, 6s CS, 4s CS, the relative level of 4s CS with respect to 6s CS (e.g. the ratio 4s CS/6s CS or the inverse ratio 6s CS/4s CS), the relative level of 6s CS with respect to 0s CS (e.g. the ratio 6s CS/0s CS or the inverse ratio 0s CS/6s CS) or the relative level of 4s CS with respect to 0s CS (e.g. the ratio 4s CS/0s CS or the inverse ratio 0s CS/4s CS).

18. The method of any one of claim 1 to 5 or 17, wherein said body fluid sample is blood and wherein an alteration in the level of one or more of the GAG properties Charge CS, Total CS, Total HS, 0s CS, 2s CS, 6s CS, 4s CS, the relative level of 4s CS with respect to 6s CS (e.g. the ratio 4s CS/6s CS or the inverse ratio 6s CS/4s CS), the relative level of 6s CS with respect to 0s CS (e.g. the ratio 6s CS/0s CS or the inverse ratio 0s CS/6s CS) or the relative level of 4s CS with respect to 0s CS (e.g. the ratio 4s CS/0s CS or the inverse ratio 0s CS/4s CS), in comparison to a control level, is indicative of cancer in said subject.

19. The method of any one of claim 1 to 5, 17 or 18, wherein said body fluid sample is blood and wherein an increase in the level of one or more of the GAG properties Charge CS, Total CS, Total HS, 6s CS, 4s CS, the ratio 6s CS/0s CS or the ratio 4s CS/0s CS, in comparison to a control level, is indicative of cancer in said subject.

20. The method of any one of claim 1 to 5, 17 or 18, wherein said body fluid sample is blood and wherein a decrease in the level of one or more of the GAG properties 0s CS, 2s CS or the ratio 4s CS/6s CS, in comparison to a control level, is indicative of cancer in said subject.

21. The method of any one of claims 1 to 15, wherein said cancer is prostate cancer.

22. The method of any one of claim 1 to 5 or 16, wherein said determination of the chemical composition comprises determining the level in the sample of one or more of the GAG properties selected from the group consisting of: 4s CS, 0s CS, Total HA and Ns2s HS.

23. The method of any one of claim 1 to 5 or 16 or 17, wherein said body fluid sample is a blood sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or both of the GAG properties 4s CS and 0s CS.

24. The method of any one of claim 1 to 5 or 16 or 17, wherein said body fluid sample is a urine sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties 4s CS, the total concentration of HA and Ns2s HS.

25. The method of any one of claim 1 to 5 or 16 or 17, wherein said body fluid sample is a blood sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 0s CS, 2s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, charge CS, Total CS, the relative level of 4s CS with respect to 6s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Tris HS, Ns2s HS, 2s6s HS, 6s HS and Total HS.

26. The method of any one of claim 1 to 5 or 16, 17, 18 or 20, wherein said body fluid sample is a blood sample and wherein an increase in the level of one or more or all of the GAG properties selected from the group consisting of 2s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, charge CS, the ratio 6s CS/4s CS, the ratio 6s CS/0s CS, the the ratio 4s CS/0s CS, Tris HS, Ns2s HS, 6s HS and Total HS, in comparison to a control level, is indicative of prostate cancer in said subject.

27. The method of any one of claim 1 to 5 or 16, 17, 18 or 20, wherein said body fluid sample is a blood sample and wherein a decrease in the level of one or more or all of the GAG properties selected from the group consisting of 0s CS, the ratio 4s CS/6s, Total CS and 2s6s HS, in comparison to a control level, is indicative of prostate cancer in said subject.

28. The method of any one of claim 1 to 5 or 16, 17 or 19, wherein said body fluid sample is a urine sample and wherein said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 4s CS, 2s6s CS, 2s4s CS, Tris CS, Total CS, the relative level of 4s CS with respect to 6s CS, the relative level of 4s CS with respect to 0s CS, charge HS, HS tot, 2s HS, Ns2s HS, Ns6s HS, Tris HS and Total HA.

29. The method of any one of claim 1 to 5 or 16, 17, 19 or 23, wherein said body fluid sample is a urine sample and wherein an increase in the level of one or more or all of the GAG properties selected from the group consisting of 2s6s CS, 2s4s CS, Tris CS, Total CS, the ratio 6s CS/4s CS, the ratio 0s CS/4s CS, charge HS, Total HS, Ns2s HS, Ns6s HS, Tris HS and Total HA, in comparison to a control level, is indicative of prostate cancer in said subject.

30. The method of any one of claim 1 to 5 or 16, 17, 19, 23 or 24, wherein said body fluid sample is a urine sample and wherein a decrease in the level of one or more or all of the GAG properties selected from the group consisting of 4s CS, the ratio 4s CS/6s CS, the ratio 4s CS/0s CS and 2s HS, in comparison to a control level, is indicative of prostate cancer in said subject.

31. The method of any one of claim 1 to 5 or 16, wherein said body fluid is blood and said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 0s CS, 2s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, charge CS, Total CS, the relative level of 4s CS with respect to 6s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Tris HS, Ns2s HS, 2s6s HS, 6s HS and Total HS; or wherein said body fluid is urine and said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 2s6s CS, 2s4s CS, Tris CS, charge HS, 2s HS, Ns2s HS, Ns6s HS and Tris HS.

32. The method of any one of claim 1 to 5 or 16, wherein said body fluid is blood and said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 2s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, charge CS, Tris HS, Ns2s HS, 2s6s HS or 6s HS; or wherein said body fluid is urine and said determination of the chemical composition comprises determining the level in the sample of one or more or all of the GAG properties selected from the group consisting of 2s6s CS, 2s4s CS, Tris CS, charge HS, 2s HS, Ns2s HS, Ns6s HS and Tris HS.

33. The method of any one of claims 1 to 5, wherein said cancer is melanoma and said method comprises determining the level in the sample of one or more of the GAG properties 6s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, Total CS, 2s HS, 0s HS and Total HS.

34. The method of any one of claim 1 to 5 or 33, wherein said cancer is melanoma and wherein an alteration in the level of one or more of the GAG properties 6s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, Total CS, 2s HS, 0s HS and Total HS, in comparison to a control level, is indicative of melanoma in said subject.

35. The method of any one of claim 1 to 5, 33 or 34, wherein said cancer is melanoma and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s6s CS, the ratio 6s CS/0s CS, Total CS and Total HS; and/or a decrease in the level of one or more of the GAG properties: the ratio 4s CS/6s CS, 2s HS and 0s HS, in comparison to a control level, is indicative of melanoma in said subject.

36. The method of any one of claims 1 to 5, wherein said cancer is colon cancer or rectum cancer and said method comprises determining the level in the sample of one or more of the GAG properties 2s CS, 6s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, Total CS, Ns6s HS, Ns2s HS, 2s6s HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS.

37. The method of any one of claim 1 to 5 or 36, wherein said cancer is colon cancer or rectum cancer and wherein an alteration in the level of one or more of the GAG properties 2s CS, 6s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, Total CS, Ns6s HS, Ns2s HS, 2s6s HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS, in comparison to a control level, is indicative of colon cancer or rectum cancer in said subject.

38. The method of any one of claim 1 to 5, 36 or 37, wherein said cancer is colon cancer or rectum cancer and wherein an increase in the level of one or more of the GAG properties: 2s CS, 6s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the ratio 6s CS/0s CS, Total CS, 6s HS, 2s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: the ratio 4s CS/6s CS, Ns6s HS, Ns2s HS, 2s6s HS and 0s HS, in comparison to a control level, is indicative of colon cancer or rectum cancer in said subject.

39. The method of any one of claims 1 to 5, wherein said cancer is a neuroendocrine tumour, preferably a gastrointestinal neuroendocrine tumour, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Ns6s HS, Ns HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS.

40. The method of any one of claim 1 to 5 or 39, wherein said cancer is a neuroendocrine tumour, preferably a gastrointestinal neuroendocrine tumour, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Ns6s HS, Ns HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS, in comparison to a control level, is indicative of a neuroendocrine tumour, preferably a gastrointestinal neuroendocrine tumour, in said subject.

41. The method of any one of claim 1 to 5, 39 or 40, wherein said cancer is a neuroendocrine tumour, preferably a gastrointestinal neuroendocrine tumour, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Ns HS, 6s HS, 2s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS, Ns6s HS and 0s HS, in comparison to a control level, is indicative of a neuroendocrine tumour, preferably a gastrointestinal neuroendocrine tumour, in said subject.

42. The method of any one of claims 1 to 5, wherein said cancer is blood cancer, preferably chronic lymphoid leukaemia, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns6s HS, Ns HS, 2s6s HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS.

43. The method of any one of claim 1 to 5 or 42, wherein said cancer is blood cancer, preferably chronic lymphoid leukaemia, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns6s HS, Ns HS, 2s6s HS, 6s HS, 2s HS, 0s HS, Charge HS and Total HS, in comparison to a control level, is indicative of blood cancer, preferably chronic lymphoid leukaemia, in said subject.

44. The method of any one of claim 1 to 5, 42 or 43, wherein said cancer is blood cancer, preferably chronic lymphoid leukaemia, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Ns HS, 6s HS, 2s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS, Ns6s HS, 2s6s HS and 0s HS, in comparison to a control level, is indicative of blood cancer, preferably chronic lymphoid leukaemia, in said subject.

45. The method of any one of claims 1 to 5, wherein said cancer is bladder cancer and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 2s CS, 6s CS, Tris CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Tris HS, Ns2s HS, Ns HS, 2s HS and Total HS.

46. The method of any one of claim 1 to 5 or 45, wherein said cancer is bladder cancer and wherein an alteration in the level of one or more of the GAG properties 0s CS, 2s CS, 6s CS, Tris CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Tris HS, Ns2s HS, Ns HS, 2s HS and Total HS, in comparison to a control level, is indicative of bladder cancer in said subject.

47. The method of any one of claim 1 to 5, 45 or 46, wherein said cancer is bladder cancer and wherein an increase in the level of one or more of the GAG properties: 2s CS, 6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Tris HS, Ns2s HS, 2s HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, Tris CS and Ns HS, in comparison to a control level, is indicative of bladder cancer in said subject.

48. The method of any one of claims 1 to 5, wherein said cancer is breast cancer and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns2s HS, Ns HS, 6s HS, 0s HS and Total HS.

49. The method of any one of claim 1 to 5 or 48, wherein said cancer is breast cancer and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns2s HS, Ns HS, 6s HS, 0s HS and Total HS, in comparison to a control level, is indicative of breast cancer in said subject.

50. The method of any one of claim 1 to 5, 48 or 49, wherein said cancer is breast cancer and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 4s6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Ns HS, 6s HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS, Ns2s HS and 0s HS, in comparison to a control level, is indicative of breast cancer in said subject.

51. The method of any one of claims 1 to 5, wherein said cancer is ovarian cancer and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Total HS and Total HA.

52. The method of any one of claim 1 to 5 or 51, wherein said cancer is ovarian cancer and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Total HS and Total HA, in comparison to a control level, is indicative of ovarian cancer in said subject.

53. The method of any one of claim 1 to 5, 51 or 52, wherein said cancer is ovarian cancer and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Total HA, and Total HS ; and/or a decrease in the level of one or both of the GAG properties: 0s CS and the ratio 4s CS/6s CS, in comparison to a control level, is indicative of ovarian cancer in said subject.

54. The method of any one of claims 1 to 5, wherein said cancer is uterine cancer, preferably endometrial cancer, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Total HS and Total HA.

55. The method of any one of claim 1 to 5 or 54, wherein said cancer is uterine cancer, preferably endometrial cancer, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Total HS and Total HA, in comparison to a control level, is indicative of uterine cancer, preferably endometrial cancer, in said subject.

56. The method of any one of claim 1 to 5, 54 or 55, wherein said cancer is uterine cancer, preferably endometrial cancer, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s4s CS, 4s6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Total HA and Total HS; and/or a decrease in the level of one or both of the GAG properties: 0s CS and the ratio 4s CS/6s CS, in comparison to a control level, is indicative of uterine cancer, preferably endometrial cancer, in said subject.

57. The method of any one of claims 1 to 5, wherein said cancer is uterine cancer, preferably cervical cancer, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns2s HS, 6s HS, Total HS and Total HA.

58. The method of any one of claim 1 to 5 or 57, wherein said cancer is uterine cancer, preferably cervical cancer, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns2s HS, 6s HS, Total HS and Total HA, in comparison to a control level, is indicative of uterine cancer, preferably cervical cancer, in said subject.

59. The method of any one of claim 1 to 5, 57 or 58, wherein said cancer is uterine cancer, preferably cervical cancer, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s4s CS, 4s6s CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Total HA, Ns2s HS, 6s HS and Total HS; and/or a decrease in the level of one or both of the GAG properties: 0s CS and the ratio 4s CS/6s CS, in comparison to a control level, is indicative of uterine cancer, preferably cervical cancer, in said subject.

60. The method of any one of claims 1 to 5, wherein said cancer is blood cancer, preferably non-Hodgkin's lymphoma, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Tris HS, Ns2s HS, Ns HS, 6s HS, 0s HS, Charge HS and Total HS.

61. The method of any one of claim 1 to 5 or 61, wherein said cancer is blood cancer, preferably non-Hodgkin's lymphoma, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Tris HS, Ns2s HS, Ns HS, 6s HS, 0s HS, Charge HS and Total HS, in comparison to a control level, is indicative of blood cancer, preferably non-Hodgkin's lymphoma, in said subject.

62. The method of any one of claim 1 to 5, 60 or 61, wherein said cancer is blood cancer, preferably non-Hodgkin's lymphoma, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s6s CS, 2s4s CS, 4s6s CS, Tris CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Tris HS, Ns2s HS, Ns HS, 6s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS and 0s HS, in comparison to a control level, is indicative of blood cancer, preferably non-Hodgkin's lymphoma, in said subject. 30

63. The method of any one of claims 1 to 5, wherein said cancer is brain cancer, preferably diffuse glioma, and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns HS, 6s HS, 0s HS, Charge HS, Total HS and Total HA.

64. The method of any one of claim 1 to 5 or 63, wherein said cancer is brain cancer, preferably diffuse glioma, and wherein an alteration in the level of one or more of the GAG properties 0s CS, 6s CS, 4s CS, 2s4s CS, 4s6s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns HS, 6s HS, 0s HS, Charge HS, Total HS and Total HA, in comparison to a control level, is indicative of brain cancer, preferably diffuse glioma, in said subject.

65. The method of any one of claim 1 to 5, 63 or 64, wherein said cancer is brain cancer, preferably diffuse glioma, and wherein an increase in the level of one or more of the GAG properties: 6s CS, 4s CS, 2s4s CS, 4s6s CS, Tris CS, the ratio 6s CS/0s CS, the ratio 4s CS/0s CS, Charge CS, Total CS, Total HA, Ns HS, 6s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS and 0s HS, in comparison to a control level, is indicative of brain cancer, preferably diffuse glioma, in said subject.

66. The method of any one of claims 1 to 5, wherein said cancer is lung cancer and said method comprises determining the level in the sample of one or more of the GAG properties 0s CS, 2s CS, 6s CS, 4s CS, 2s4s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns6s HS, Ns2s HS, 6s HS, 0s HS, Charge HS and Total HS.

67. The method of any one of claim 1 to 5 or 66, wherein said cancer is lung cancer and wherein an alteration in the level of one or more of the GAG properties: 0s CS, 2s CS, 6s CS, 4s CS, 2s4s CS, Tris CS, the relative level of 6s CS with respect to 4s CS, the relative level of 6s CS with respect to 0s CS, the relative level of 4s CS with respect to 0s CS, Charge CS, Total CS, Ns6s HS, Ns2s HS, 6s HS, 0s HS, Charge HS and Total HS, in comparison to a control level, is indicative of lung cancer in said subject.

68. The method of any one of claim 1 to 5, 66 or 67, wherein said cancer is lung cancer and wherein an increase in the level of one or more of the GAG properties: 2s CS, 6s CS, 4s CS, 2s4s CS, Tris CS, the ratio 6s CS/0s CS, ratio 4s CS/0s CS, Charge CS, Total CS, 6s HS, Charge HS and Total HS; and/or a decrease in the level of one or more of the GAG properties: 0s CS, the ratio 4s CS/6s CS, Ns6s HS, Ns2s HS and 0s HS, in comparison to a control level, is indicative of lung cancer in said subject.

69. The method of any one of claims 1 to 68, wherein said method comprises determining the level of more than one of said GAG properties.

70. The method of any one of claims 1 to 69, wherein said method comprises determining the level of two or more, three or more, four or more, or all, of said GAG properties.

71. The method of any one of claims 1 to 70, wherein said method is used for diagnosing said cancer.

72. The method of any one of claims 1 to 70, wherein said method is used for the prognosis of said cancer, for monitoring subjects at risk of the occurrence of said cancer, for monitoring the progression of said cancer in a subject, for determining the clinical severity of said cancer, for predicting the response of a subject to therapy or surgery for said cancer, for determining the efficacy of a therapeutic or surgical regime being used to treat said cancer, for detecting the recurrence or relapse of said cancer, for patient selection or treatment selection or for distinguishing small masses suspicious of said cancer from other non-malignant diseases.

73. The method of any one of claims 1 to 32 or 69 to 72, wherein said prostate cancer is prostate adenocarcinoma.

74. The method of any one of claims 1 to 73, wherein said subject is a subject at risk of developing said cancer, or at risk of the occurrence of said cancer, or is a subject having or suspected of having said cancer.

75. The method of any one of claims 1 to 6, 8, 9, 11, 14, 15, 21, 22, 24 or 28 to 74, wherein said body fluid sample is urine.

76. The method of any one of claims 1 to 7, 9, 10, 12, 13, 16 to 23. 25 to 27 or 31 to 74, wherein said body fluid sample is blood.

77. The method of any one of claims 1 to 76, wherein said level or chemical composition of said GAG or GAG property is determined by electrophoresis or by HPLC and mass spectrometry.

78. The method of claim 77, wherein said electrophoresis is capillary electrophoresis, preferably capillary electophoresis with laser-induced fluorescence detection, and/or wherein HPLC is combined with electrospray ionization-mass spectrometry.

79. The method of any one of claims 3 to 78, wherein the levels of one or more of the specific sulfated or unsulfated forms of CS or HS disaccharides are determined, and wherein the GAGs are subjected to a processing step to obtain the disaccharide units for analysis.

80. The method of any one of claims 1 to 79, wherein said method further comprises a step of treating said cancer by therapy or surgery.

81. A method of screening for cancer in a subject, said method comprising determining the level in a sample of an expression product of one or more genes selected from the group consisting of: UST, CHST14, CHST13, CHSY1, DSE, CHSY3, CHST11, CHST15, CHPF2, CSGALNACT2, CHST12, CSGALNACT1, CHST7, CHPF, CHST3, HPSE, HGSNAT, HYAL4, GALNS, GLB1, GNS, GUSB, HEXA, HEXB, HYAL1, IDS, IDUA, ARSB, NAGLU, HPSE2, SGSH, SPAM1, HYAL3, HYAL2, HS3ST3B1, HS3ST3A1, B4GALT7, B3GALT6, B3GAT2, B3GAT3, B3GAT1, XYLT1, XYLT2, EXT1, EXT2, EXTL1, EXTL2, EXTL3, HS3ST5, GLCE, HS6ST3, NDST1, NDST4, HS6ST2, NDST3, HS6ST1, HS2ST1, HS3ST2, HS3ST1 and NDST2; wherein said sample has been obtained from said subject; and wherein an altered level in said sample of the expression product of one or more of said genes in comparison to a control level is indicative of cancer in said subject.

82. The method of claim 81, wherein said cancer is selected from the group consisting of prostate cancer, thyroid cancer, colon cancer, rectum cancer, lung cancer, uterine cancer, breast cancer, pancreatic cancer, bladder cancer, liver cancer, bile duct cancer, stomach cancer, oesophageal cancer, head and neck cancer, brain cancer, blood cancer, ovarian cancer, skin cancer and kidney cancer.

83. The method of claim 81 or claim 82, wherein said cancer is selected from the group consisting of prostate cancer, thyroid cancer, colon cancer, rectum cancer, lung cancer, uterine cancer, breast cancer, pancreatic cancer, bladder cancer, liver cancer, bile duct cancer, stomach cancer, oesophageal cancer, head and neck cancer, brain cancer, blood cancer, ovarian cancer and skin cancer.

Description

[0673] The invention will be further described with reference to the following non-limiting Example with reference to the following drawings in which:

[0674] FIG. 1. Principal component analysis of the GAG profiles in the blood of healthy (light grey), PRAD (dark grey), or clear cell RCC (black) subjects. Subjects belonging to a given group are represented as points connected to the geometrical center of the ellipsis representing that group. PRAD: Prostate Cancer. RCC: Renal Cell Carcinoma.

[0675] FIG. 2. Principal component analysis of the GAG profiles in the urine of healthy (light grey), PRAD (dark grey), or clear cell RCC (black) subjects. Subjects belonging to a given group are represented as points connected to the geometrical center of the ellipsis representing that group. PRAD: Prostate Cancer. RCC: Renal Cell Carcinoma.

[0676] FIG. 3. Principal component analysis of the GAG profiles in the blood and urine combined of healthy (light grey), PRAD (dark grey), or clear cell RCC (black) subjects. Subjects belonging to a given group are represented as points connected to the geometrical center of the ellipsis representing that group. PRAD: Prostate Cancer. RCC: Renal Cell Carcinoma.

[0677] FIG. 4. Boxplots of blood, urine, and combined scores in healthy vs. PRAD subjects. Individual subject scores are represented as dots. The horizontal lines identify the optimal cut-off score for which the highest accuracy in distinguishing PRAD from healthy individuals was achieved for each type of score. PRAD: Prostate Cancer.

[0678] FIG. 5. ROC curves for the classification of subjects as PRAD versus healthy based on the blood (top), urine (middle), or combined (bottom) score. The points overlaid on each curve represent the optimal cut-off score at which the classification of subjects as PRAD versus healthy based on the different scores is most accurate (the two-dimensional coordinates of that point being the specificity and the sensitivity). PRAD: Prostate Cancer.

[0679] FIG. 6. Boxplots of PRAD combined scores in healthy vs. PRAD vs. clear cell RCC subjects. The horizontal lines identify the optimal cut-off score previously detected to maximize the accuracy in distinguishing PRAD from healthy individuals. PRAD: Prostate Cancer. RCC: Renal Cell Carcinoma.

[0680] FIG. 7. Boxplots of RCC blood scores in healthy vs. PRAD vs. clear cell RCC subjects. The horizontal lines identify the optimal cut-off score previously detected to maximize the accuracy in distinguishing RCC from healthy individuals. PRAD: Prostate Cancer. RCC: Renal Cell Carcinoma.

[0681] FIG. 8. Ranking of the GAG properties in terms of decreased accuracy of a random forest classifier when the property is omitted (the decrease in accuracy is measured as the mean decrease in the Gini coefficient). The classifier was trained on GAG properties of matched blood and urine samples. Only the 30 best properties are shown.

[0682] FIG. 9(A/B/C) GAG properties in the blood that were significantly altered in cancer as opposed to healthy volunteers but not due to cancer type specific effects. Triangle and circle samples in the boxplot representing the cancer population define two different types of cancer. (A: NsHS; B: 6sCS and Charge CS; C: 0sCS and 4s/6sCS).

[0683] FIG. 10(A/B/C) GAG properties in the urine that were significantly altered in cancer as opposed to healthy volunteers but not due to cancer type specific effects. Triangle and circle samples in the boxplot representing the cancer population define two different types of cancer. (A: 0sHS and Charge HS; B: 4sCS and 6s/0sCS; C:4s/0sCS).

[0684] FIG. 11Receiver operating characteristic (ROC) curve for the classification of samples into casessubjects with present diagnosis of cancervs. controlshealthy subject or with former diagnosis of cancer and currently with no evidence of diseaseusing a multilayer perceptron classifier trained on 66% of samples' GAG profiles (left) and tested on the remaining 33% of samples GAG profiles (right) as measured in the blood (top), urine (middle), or both blood and urine (bottom).

[0685] FIG. 12GAG scores based on blood GAG property measurements in casessubjects with present diagnosis of cancervs. controlshealthy subject or with former diagnosis of cancer and currently with no evidence of disease. The horizontal line identifies the optimal cut-off score for which the highest accuracy in distinguishing cases from controls was achieved. One case data point was omitted because the GAG scores was greater than 3. Key: squaresrenal cell carcinoma; trianglesprostate cancer; circleshealthy subjects. Note that a control may be represented with a cancer symbol if the present status at the time of sampling was no evidence of disease.

[0686] FIG. 13GAG scores based on urine GAG property measurements in casessubjects with present diagnosis of cancervs. controlshealthy subject or with former diagnosis of cancer and currently with no evidence of disease. The horizontal line identifies the optimal cut-off score for which the highest accuracy in distinguishing cases from controls was achieved. Key: squaresrenal cell carcinoma; trianglesprostate cancer; circleshealthy subjects. Note that a control may be represented with a cancer symbol if the present status at the time of sampling was no evidence of disease.

[0687] FIG. 14GAG scores based on blood and urine GAG property measurements in casessubjects with present diagnosis of cancervs. controlshealthy subject or with former diagnosis of cancer and currently with no evidence of disease. The horizontal line identifies the optimal cut-off score for which the highest accuracy in distinguishing cases from controls was achieved. Key: squaresrenal cell carcinoma; trianglesprostate cancer; circleshealthy subjects. Note that a control may be represented with a cancer symbol if the present status at the time of sampling was no evidence of disease.

[0688] FIG. 15. ROC curves for the classification of subjects as cases versus controls based on the blood (top), urine (middle), or combined (bottom) score. The points overlaid on each curve represent the optimal cut-off score at which the classification of subjects as case versus control based on the different scores is most accurate (the two-dimensional coordinates of that point being the specificity and the sensitivity). The area-under-the-curve (AUC) of each ROC curve is also reported.

[0689] FIG. 16. Regulation of glycosaminoglycan metabolism in 22 subtypes of cancer vs. matched normal tissue of origin. Heat-map of the direction of regulation of each gene associated with glycosaminoglycan metabolism (rows) in each cancer subtype (columns) compared to the matched normal tissue of origin. Grey entries indicate gene expression level down-regulation, black entries indicate gene expression level up-regulation, while white entries indicate no significant gene expression level difference with normal samples or no data available. Legend: MMMalignant melanoma; PRADProstate adenocarcinoma; THCAThyroid carcinoma; PAADPancreatic adenocarcinoma; COADColon adenocarcinoma; READRectal adenocarcinoma; UCECUterine corpus endometrial carcinoma; BRCABreast carcinoma; LUADLung adenocarcinoma; BLADUrothelial bladder carcinoma; LUSCLung squamous cell carcinoma; LIHCLiver hepatocellular carcinoma; CHOLCholangiocarcinoma; GBMGlioblastoma multiforme; STADStomach adenocarcinoma; HNSCHead and neck squamous cell carcinoma; ESCAEsophogael carcinoma; OVSASerous ovarian adenocarcinoma; PTCLPeripheral T-cell Lymphoma; KIRCClear cell renal cell carcinoma; KICHChromophobe renal cell carcinoma; KIRPPapillary renal cell carcinoma.

[0690] FIG. 17. Boxplots of M GAG blood scores in healthy vs. melanoma. The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing melanoma from healthy individuals for this score.

[0691] FIG. 18. Boxplots of M GAG blood scores calculated using an alternative formula in healthy vs. melanoma. The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing melanoma from healthy individuals for this alternative score.

[0692] FIG. 19. Boxplots of CRC GAG blood scores in healthy vs. colorectal cancer (CRC). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing CRC from healthy individuals for this score.

[0693] FIG. 20. Boxplots of GNET GAG blood scores in healthy vs. gastrointestinal neuroendocrine tumors (GNET). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing GNET from healthy individuals for this score.

[0694] FIG. 21. Boxplots of CLL GAG blood scores in healthy vs. chronic lymphoid leukemia (CLL). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing CLL from healthy individuals for this score.

[0695] FIG. 22. Boxplots of CLL GAG blood scores calculated using an alternative formula in healthy vs. chronic lymphoid leukemia (CLL). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing CLL from healthy individuals for this alternative score.

[0696] FIG. 23. Boxplots of BCa GAG urine scores in healthy vs. bladder cancer (BCa). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing BCa from healthy individuals for this score.

[0697] FIG. 24. Boxplots of BC GAG blood scores in healthy vs. breast cancer (BC). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing BC from healthy individuals for this score.

[0698] FIG. 25. Boxplots of OV GAG blood scores in healthy vs. ovarian cancer (OV). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing OV from healthy individuals for this score.

[0699] FIG. 26. Boxplots of EC GAG blood scores in healthy vs. endometrial cancer (EC). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing EC from healthy individuals for this score.

[0700] FIG. 27. Boxplots of CST GAG blood scores in healthy vs. cervical cancer (CST). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing CST from healthy individuals for this score.

[0701] FIG. 28. Boxplots of NHL GAG blood scores in healthy vs. non-Hodgkin lymphoma (NHL). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing NHL from healthy individuals for this score.

[0702] FIG. 29. Boxplots of DG GAG blood scores in healthy vs. diffuse glioma (DG), grouped by subtype. The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing DG from healthy individuals for this score.

[0703] FIG. 30. Boxplots of LC GAG blood scores in healthy vs. lung cancer (LC). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing LC from healthy individuals for this score.

[0704] FIG. 31. Boxplots of LC GAG blood scores calculated using an alternative formula in healthy vs. lung cancer (LC). The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing LC from healthy individuals for this alternative score.

[0705] FIG. 32. Boxplots of CS charge and total CS and HA concentration in each cancer type vs. healthy subjects. Key: breast cancerBC, colorectal cancerCRC, cervical cancer CST, diffuse gliomaDG, endometrial cancerEC, gastrointestinal neuroendocrine tumorsGNET, healthyH, lymphoid leukemiaLL, Non-Hodgkin's lymphomaNHL, lung cancerNSCLC, ovarian cancerOV, prostate cancerPCa.

[0706] FIG. 33. Boxplots of HS charge and total HS in each cancer type vs. healthy subjects. Key as in FIG. 32.

[0707] FIG. 34. Mean value of each property in CS composition in each cancer type group and in the healthy subjects' group. Mean values for each CS composition property belonging to the same group are connected by a distinct line. Key as in FIG. 32.

[0708] FIG. 35. Mean value of each property in HS composition in each cancer type group and in the healthy subjects' group. Mean values for each HS composition property belonging to the same group are connected by a distinct line. Key as in FIG. 32.

[0709] FIG. 36. Boxplots of cancer GAG blood scores calculated using an alternative formula in healthy vs. cancer grouped by type. Key as FIG. 32. The horizontal line identifies the optimal cut-off score that maximizes the accuracy in distinguishing cancer from healthy individuals for this alternative score. On the right, the ROC corresponding to the cancer versus healthy classification using this alternative cancer GAG blood score.

EXAMPLE 1

[0710] Introduction

[0711] The number of prostate cancer cases is predicted to increase substantially in the near future. The rising prostate cancer population determines an urgent need to improve the current diagnostics landscape for prostate cancer. In particular, affordable and practical tools for prostate cancer diagnostics are needed to assist healthcare professionals in the early detection of prostate cancer, which typically correlates with more favorable clinical outcomes, or to guide treatment of current prostate cancer patients. Circulating biomarkers are molecules that can be measured in accessible body fluids of individuals, e.g. blood or urine, and whose levels are useful to assist in the diagnosis and/or prognosis and/or prediction of response to treatment. An example of a widely used biomarker is the prostate-specific antigen (PSA) for prostate cancer, but the clinical value of this biomarker for diagnosing prostate cancer is highly debated.

[0712] Emerging cancer biomarkers are circulating nucleic acids directly derived from cancer cells. These nucleic acids can be found for example in circulating free DNA, circulating small-RNA, circulating tumor cells, or extracellular micro-vesicles (including exosomes) containing small-RNA, mRNA and DNA. The minimally invasive technology for detection of such molecular biomarkers without the need for costly or invasive procedures is defined as liquid biopsy. Recently, the definition of liquid biopsy was extended to encompass other macromolecular classes of cancer biomarkers, for example proteins transported within circulating exosomes. It has previously been demonstrated that even circulating metabolites could serve as accurate molecular biomarkers for renal cell carcinoma (RCC), the most common form of kidney cancer (Gatto et al., 2016). In particular, quantitative profiling of glycosaminoglycans (GAGs) in a subjects' blood and/or urine could be scored computationally to derived GAG scores with demonstrated diagnostic value for RCC (Gatto et al., 2016).

[0713] In this study we measured circulating GAG levels in subjects presenting with prostate cancer.

[0714] Material and Methods

[0715] Blood and Urine Sample Collection

[0716] 28 matched urine and blood samples were collected from 14 patients with PRAD.

[0717] Whole blood tubes were centrifuged (2,500 g for 15 minutes at 4 C.) and the serum extracted and collected in a separate tube. Blood GAG measurements in patients with PRAD were done on serum samples. Urine samples were collected in polypropilene tubes. The samples were stored at 80 C. until they were shipped for analysis in dry ice.

[0718] Urine samples and blood (plasma) samples from healthy subjects and RCC subjects are as per Gatto et al., 2016.

[0719] Serum and plasma GAGs correspond to blood GAGs. For the purpose of this study, serum and plasma are therefore referred to collectively as blood.

[0720] Glycosaminoglycan Profile Determination

[0721] Sample preparation including extraction and purification steps were performed as previously described by Volpi and Maccari 2005.sup.1 and 2 and Coppa et al., 2011), while sample GAGs separation and quantification were performed as described in Volpi et al., 2014; Volpi and Linhardt, 2010.

[0722] Briefly, to extract the GAGs, 500 l of sample was lyophilized, reconstituted with 1 ml of a 20-mM TRIS-Cl buffer pH 7.4 and treated with protease (Proteinase K from Tritirachium album [E.C. 3.4.21.64], >500 units ml_1 from Sigma-Aldrich) at 60 C. for 12 h. After boiling for 10 min, centrifugation and filtration on 0.45-m filters, the filtrate was lyophilized. The powder was dissolved in 1 ml of distilled water by prolonged mixing. After centrifugation at 5000 g for 15 min, 0.2 ml of 20% trichloro-acetic acid was added to the supernatant. After 2 h at 4 C., the mixture was centrifuged at 5000 g for 15 min, and the supernatant was recovered and lyophilized. After solubilization in 0.4 ml of bidistilled water and centrifugation at 10 000 g for 10 min, the supernatant was collected and further analysed. To purify the GAGs, after reconstitution with 500 l 10 mM NaCl, sample GAGs were further purified on anion-exchange resin (QAE Sephadex A-25). After centrifugation at 10,000g for 5 min, the supernatant was applied to a column (1 cm4 cm) packed with about 3 ml of resin previously equilibrated with 10 mM NaCl. After washing the resin with 20 ml of 10 mM NaCl, 10 ml of 2.5 M NaCl were added. Fifty milliliters of ethanol were added to the eluate (10 ml) and stored at 20 C. for 24 h. After centrifugation at 5000g for 15 min, the pellet was dried at 60 C. for 12 h. After reconstitution with 80 l of 50 mM ammonium acetate pH 8.0, the material was treated with 20 l of chondroitinase ABC at 37 C. for 12 h. To separate and quantify the GAGs, after boiling for 5 min, the samples were injected in HPLC with both post-column derivatization and fluorimetric detection and on-line electrospray ionization mass spectrometry (ESI-MS).

[0723] Nineteen independent GAG properties were measured in each sample (either blood or urine): CS concentration (in g/mL), HS concentration (in g/mL), HA concentration (in g/mL), and mass fractions of disaccharide composition for both CS and HS. In addition, five dependent GAG properties were calculated from these measurements: the CS and HS charge; and the following CS ratio of mass fractions: 4s/6s, 6s/0s, and 4s/0s. The 24 GAG properties may be collectively referred to as GAG profile.

[0724] The overall similarity in the GAG profile between subjects belonging to the three groups (prostate cancer vs. renal cell carcinoma vs. healthy) was further examined using principal component analysis (PCA) by comparing GAG properties measured either solely in the blood or in the urine or in both fluids. Principal component analysis was implemented using R-package ade4 (centering was performed by the mean) (Dray and Dufour, 2007).

[0725] Bio-Marker Design

[0726] The statistical significance of changes for each GAG property between the two groups (prostate cancer vs. healthy) was assessed by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation under the following assumptions: GAG property values are sampled from a t-distribution of unknown and to be estimated normality (i.e. degrees of freedom); high uncertainty on the prior distributions; the marginal distribution is well approximated by a Markov chain Monte Carlo sampling with no thinning and chain length equal to 100'000. The estimation was performed using BEST R-package (the above assumptions are reflected by the default parameters). Bayesian estimation was preferred over the widely used t-test since it provides a robust and reliable estimation of mean difference even under uncertainty of the underlying score distribution for the two groups (that is the case when the number of samples is limited).

[0727] To verify whether changes in the GAG profile in PRAD patients could be condensed into GAG scores, that can be used to differentiate PRAD patients from healthy subjects, and therefore would be suitable as a cancer diagnostic, we utilized Lasso penalized logistic regression (Tibshirani, 1996) with leave-one-out cross-validation to select robust GAG properties that are most predictive of the clinical outcome (i.e., PRAD compared to healthy subject). The score was subsequently designed as a ratio, where the numerator is the sum of the properties associated with PRAD and the denominator is the sum of the properties associated with the healthy state. Each term was normalized using the regression coefficients.

[0728] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was serum or plasma are referred to as GAG blood scores.

[0729] Accuracy Metrics

[0730] For each marker (blood, urine, or combined), we evaluated its performance in the binary classification of a sample as either prostate cancer or healthy at varying threshold scores by deriving the receiver-operating-characteristic (ROC) curves. We measured the accuracy of each marker as the area under the curve (AUC) of its ROC curve (AUC is 1 for a perfect classifier and 0.5 for a random classifier). We selected as a potential cut-off value for a given marker the score for which the accuracy was maximum, i.e. a sample whose marker score is above this cut-off value has the maximum probability of being prostate cancer and below this cut-off value has the maximum probability of being healthy. The ROC curves were calculated using the pROC R-package, while the optimal cut-off using the OptimalCutpoints R-package.

[0731] Results

[0732] We analyzed GAG profiles in prostate adenocarcinoma (PRAD). We collected a total of 28 matched urine and blood samples from 14 patients with PRAD. Clinical data for this cohort is reported in Table A. We quantified the profile of GAGs in the former samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile for both blood and/or urine samples (listed in Table B). We also compared the value of each property in the GAG profile of the PRAD group to previously measured values in the GAG profile of a group of 50 matched urine and blood samples from 25 healthy individuals. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation under the following assumptions: GAG property values are sampled from a t-distribution of unknown and to be estimated normality (i.e. degrees of freedom); high uncertainty on the prior distributions; the marginal distribution is well approximated by a Markov chain Monte Carlo sampling with no thinning and chain length equal to 100'000. The estimation was performed using BEST R-package (the above assumptions are reflected by the default parameters). Bayesian estimation was preferred over the widely used t-test since it provides a robust and reliable estimation of mean difference even under uncertainty of the underlying score distribution for the two groups (that is the case when the number of samples is limited). The results showed marked differences in several GAG properties between PRAD, compared to healthy subjects (Table B). We further examined overall similarity in the GAG profile between subjects belonging to the two groups using principal component analysis (PCA) by comparing GAG properties measured either solely in the blood (FIG. 1) or in the urine (FIG. 2) or in both fluids (FIG. 3). Principal component analysis was implemented using R-package ade4 (centering was performed by the mean) (Dray and Dufour, 2007). This analysis demonstrated that subjects belonging to the two groups clustered separately in either fluid, suggesting that marked differences exist in the profile of GAGs for the two groups. To compare how similar the GAG profile in PRAD was to that of RCC, we included in each PCA the GAG profiles from the RCC group, analyzed in our previous study (Gatto et al., 2016) and the blood GAG profiles collected from an additional cohort of 40 clear cell RCC subjects. The RCC group from our previous study included specifically clear cell RCC patients, comprising 52 subjects with blood samples, 20 subjects with urine samples, and 19 subjects with both. Remarkably, a separation was observed both between PRAD and RCC subjects, and that also was separate from healthy subjects, indicating that unique changes in the GAG profile occur in the blood and urine of PRAD subjects.

[0733] To verify whether changes in the GAG profile in PRAD patients could be condensed into GAG scores, that can be used to differentiate PRAD patients from healthy subjects, and therefore would be suitable as a prostate cancer diagnostic, we utilized Lasso penalized logistic regression (Tibshirani, 1996) with leave-one-out cross-validation to select robust GAG properties that are most predictive of the clinical outcome (i.e., PRAD compared to healthy subject). The score was subsequently designed as a ratio, where the numerator is the sum of the properties associated with PRAD and the denominator is the sum of the properties associated with the healthy state. Each term was normalized using the regression coefficients. We derived three PRAD biomarker scores, based on either blood or urine or combined measurements:

[00016] $Blood .Math. .Math. score = \frac{[4 .Math. s .Math. .Math. CS]}{\frac{4}{10} [0 .Math. .Math. s .Math. .Math. CS]}$ $Urine .Math. .Math. score = \frac{3 [HA] + \frac{6}{10} [Ns .Math. .Math. 2 .Math. s .Math. .Math. HS]}{\frac{1}{3} [4 .Math. .Math. s .Math. .Math. CS]}$ $Combined .Math. .Math. score = \frac{(Blood .Math. .Math. score + Urine .Math. .Math. score)}{2}$

where terms in brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and [HA] is the total concentration of HA (in g/mL). We subsequently calculated the three scores for each sample and observed that PRAD samples have recurrently elevated scores with respect to healthy samples (FIG. 4). We computed significant non-null mean differences in all three scores between the two groups using robust Bayesian estimation. The mean difference was equal to 0.43 for the combined score (95% HDI 0.35 to 0.52), 0.54 for the blood score (95% HDI 0.39 to 0.69), and 0.29 for the urine score (95% HDI 0.21 to 0.37). The performance of the three scores was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier) in the case of the combined score, 0.997 for the blood score, and 0.994 for the urine score (FIG. 5). We identified an optimal cut-off for each score to maximize the accuracy in the classification of PRAD subjects as opposed to healthy individuals using blood and/or urine GAG profile data. These cut-off scores were 0.63, 0.19, 0.47 for the blood, urine, and combined score respectively. Using these cut-offs, PRAD subjects could be distinguished from healthy individuals with 97.4% accuracy using the blood or urine score, and 100% accuracy using the combined score. Other metrics of accuracy were reported in Table C. Taken together, these results demonstrate that GAG profiling in accessible biofluids are suitable for diagnosing prostate cancer.

[0734] We also investigated whether scores designed to detect PRAD are not specific to RCC (renal cell carcinoma) and vice versa. Strikingly, the combined score designed to detect PRAD specifically correlated with the 14 subjects with PRAD, as opposed to 19 subjects with RCC and 25 healthy subjects (FIG. 6). By optimizing the cut-off to maximize accuracy in the detection of PRAD, at a cut-off score equal to 0.52, PRAD can be distinguished from healthy subjects or subjects with RCC with 98.3% accuracy, conferring to the combined score an overall sensitivity and specificity for PRAD equal to 100% and 97.7% respectively. For comparison, a standard PSA test to detect PRAD in men over 50 years old at average risk assuming a cut-off value equal to 4 ng/ml has typical values for sensitivity and specificity equal to 21% (51% for high grade lesions with Gleason score greater or equal to 8) and 91% respectively (Wolf et al., 2010).

[0735] Conversely, the blood score designed to detect RCC specifically identified subjects with RCC, as opposed to subjects with PRAD and healthy subjects (FIG. 7). By optimizing the cut-off to maximize accuracy in the detection of RCC as opposed to healthy subjects or subjects with PRAD, at a cut-off score equal to 0.23 for the previously derived blood score for RCC (Gatto et al., 2016), the overall accuracy also for the detection of subjects with RCC was 89.3%, achieving a sensitivity and specificity equal to 97.8% and 69.2%, respectively and an AUC =0.941. These results demonstrate that different GAG scores (or GAG profiles) can be designed to specifically detect prostate or RCC, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

[0736] Conclusions

[0737] Liquid biopsies hold promise to revolutionize prostate cancer diagnostics by providing an accurate, simple, minimally invasive, fast and/or cheaper alternative to tissue biopsies and medical imaging, currently the gold standard for prostate cancer diagnostics. In this study, we have demonstrated that an emerging liquid biopsy platform could leverage on blood and/or urine based measurements of GAG profiles. We demonstrated that there are measurable differences in circulating GAG levels and/or chemical composition between prostate cancer as opposed to healthy subjects. The GAG profiles can further be condensed into GAG scores that have great promise in terms of accuracy for prostate cancer detection or diagnosis. Furthermore, the prostate cancer GAG scores were shown to be specific to prostate cancer compared to RCC (renal cell carcinoma). The diagnostic GAG properties or scores disclosed herein could also be applied to change a number of clinical practices. For example, to monitor prostate cancer before and after surgery or drug treatment; to rule out the relapse of the disease during a longer period of time after which a patient is typically declared cured; to assess the occurrence of prostate cancer in a population at risk, such as genetically predisposed individuals or individuals presenting risk factors or individuals presenting symptoms; to ascertain whether a metastasis is due to prostate cancer; to predict recurrence or relapse in patients with early stage prostate cancer; to distinguish lesions suspicious of prostate cancer from non malignant diseases; and to screen for prostate cancer in the general population. This liquid biopsy has therefore the potential to become a crucial tool to enable secondary prevention of prostate cancer and reduce the burden caused by this disease in the general population.

REFERENCES

[0738] Coppa G V, Gabrielli O, Buzzega D, Zampini L, Galeazzi T, Maccari F, Bertino E, Volpi N (2011). Composition and structure elucidation of human milk glycosaminoglycans. Glycobiology 21, 295-303.

[0739] Dray, S., and Dufour, A. B. (2007). The ade4 package: Implementing the duality diagram for ecologists. Journal of Statistical Software 22, 1-20.

[0740] Gatto, F., Volpi, N., Nilsson, H., Nookaew, I., Maruzzo, M., Roma, A., Johansson, M. E., Stierner, U., Lundstam, S., Basso, U., et al. (2016). Glycosaminoglycan Profiling in Patients' Plasma and Urine Predicts the Occurrence of Metastatic Clear Cell Renal Cell Carcinoma. Cell Rep 15, 1822-1836.

[0741] Tibshirani, R. (1996). Regression shrinkage and selection via the Lasso. Journal of the Royal Statistical Society Series B-Methodological 58, 267-288.

[0742] Volpi N and Maccari F. (2005.sup.1). Glycosaminoglycans composition of the largefreshwater mollusc bivalve Anodonta anodonta. Biomacromolecules 6, 3174-3180.

[0743] Volpi N and Maccari F (2005.sup.2). Microdetermination of chondroitin sulfate in normal human plasma by fluorophore-assisted carbohydrate electrophore-sis (FACE). Clin Chim Acta 356, 125-133.

[0744] Volpi, N., Galeotti, F., Yang, B., and Linhardt, R. J. (2014). Analysis of glycosaminoglycan-derived, precolumn, 2-aminoacridone-labeled disaccharides with LC-fluorescence and LC-MS detection. Nature protocols 9, 541-558.

[0745] Volpi, N., and Linhardt, R. J. (2010). High-performance liquid chromatography-mass spectrometry for mapping and sequencing glycosaminoglycan-derived oligosaccharides. Nature protocols 5, 993-1004.

[0746] Wolf, A. M., Wender, R. C., Etzioni, R. B., Thompson, I. M., D'Amico, A. V., Volk, R. J., Brooks, D. D., Dash, C., Guessous, I., Andrews, K., et al. (2010). American Cancer Society guideline for the early detection of prostate cancer: update 2010. CA Cancer J Clin 60, 70-98.

TABLE-US-00001 TABLE A Clinical data for the cohorts analysed in this study. All results are presented as medians (25.sup.th, 75.sup.th percentile) or percent. Missing values were omitted. PRAD Healthy ccRCC (n = 14) (n = 25) (n = 99) Cohort characteristics Matched blood + urine samples 14 25 19 Blood samples 0 0 92 Urine samples 0 0 20 Baseline characteristics Age [years] 69 (66-71) 61 (55-64) 63 (56-71) Female 0% 62.5% 19% BMI [kg/m.sup.2] 22 (22-28) 26 (23-29) 25 (23-28) PSA level [ng/ml] 11 (9-19)

[0747] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between PRAD and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table B shows a summary of the results and shows the mean value for GAG properties as measured in PRAD vs. healthy subjects blood or urine and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table B also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00002 TABLE B Mean in Mean in Healthy Prostate ROPE % in GAG property Individuals Cancer threshold ROPE 0 s CS urine 34.9150 40.3666 2.0183 16.42 2 s CS urine 0.0852 0.1049 0.0052 12.42 6 s CS urine 21.5534 25.9494 1.2975 8.60 4 s CS urine 40.5560 30.5398 1.5270 0.01 2 s6 s CS urine 0.6060 0.8793 0.0440 0.71 2 s4 s CS urine 0.4569 0.6285 0.0314 1.30 4 s6 s CS urine 1.1849 1.3990 0.0700 14.51 Tris CS urine 0.0387 0.0974 0.0049 0.11 4 s/6 s CS urine 2.0071 1.1994 0.0600 0.00 6 s/0 s CS urine 0.7510 0.7021 0.0351 19.13 4 s/0 s CS urine 1.3096 0.8078 0.0404 0.48 Charge CS urine 0.6532 0.6289 0.0314 47.35 Total HA urine 0.0703 0.2386 0.0119 0.11 Total CS urine 0.9124 2.8992 0.1450 0.29 0 s CS blood 81.9248 68.7842 3.4392 0.00 2 s CS blood 0.0163 0.0478 0.0024 1.38 6 s CS blood 0.2109 0.8672 0.0434 0.00 4 s CS blood 17.6142 29.3080 1.4654 0.00 2 s6 s CS blood 0.0100 0.0335 0.0017 0.27 2 s4 s CS blood 0.0045 0.0269 0.0013 0.63 4 s6 s CS blood 0.0222 0.0949 0.0047 0.14 Tris CS blood 0.0051 0.0164 0.0008 0.97 4 s/6 s CS blood 69.2937 27.8682 1.3934 0.00 6 s/0 s CS blood 0.0025 0.0102 0.0005 0.00 4 s/0 s CS blood 0.2154 0.4310 0.0215 0.00 Charge CS blood 0.1805 0.3163 0.0158 0.00 Total HA blood 0.1161 0.1393 0.0070 12.59 Total CS blood 5.3857 4.0150 0.2007 4.23 Tris HS urine 0.3545 1.6758 0.0838 0.01 Ns6 s HS urine 1.1990 3.2476 0.1624 0.00 Ns2 s HS urine 1.3681 4.8767 0.2438 0.00 Ns HS urine 12.2820 13.6728 0.6836 20.50 2 s6 s HS urine 1.0458 1.6221 0.0811 6.48 6 s HS urine 6.8895 9.0528 0.4526 9.57 2 s HS urine 3.5227 2.3676 0.1184 3.32 0 s HS urine 71.8320 62.4626 3.1231 6.53 Charge HS urine 0.3329 0.5107 0.0255 0.28 Total HS urine 0.1443 0.3534 0.0177 2.49 Tris HS blood 0.3189 1.0385 0.0519 0.23 Ns6 s HS blood 1.5484 1.1908 0.0595 7.95 Ns2 s HS blood 0.6587 1.9633 0.0982 0.46 Ns HS blood 15.9138 11.2063 0.5603 5.07 2 s6 s HS blood 0.9313 0.3779 0.0189 0.40 6 s HS blood 6.8018 15.1316 0.7566 2.60 2 s HS blood 4.7854 6.0821 0.3041 12.35 0 s HS blood 65.0374 56.7356 2.8368 16.16 Charge HS blood 0.4166 0.5036 0.0252 13.88 Total HS blood 0.0433 0.1573 0.0079 0.03

TABLE-US-00003 TABLE C Accuracy metrics at optimal cut-off for the blood, urine, and combined score in distinguishing 14 PRAD vs. 25 healthy subjects. Score AUC Accuracy Sensitivity Specificity Blood 0.997 97.4% 100% 96% Urine 0.994 97.4% 100% 96% Combined 1 100% 100% 100%

EXAMPLE 2

[0748] Introduction

[0749] The number of cancer cases is predicted to increase substantially in the near future. In the US alone, the number of new cancer patients is expected to rise from 1.36 million in 2000 to almost 3.0 million in 2050. The rising cancer population determines an urgent need to improve the current diagnostics landscape for cancer. In particular, affordable and practical tools for cancer diagnostics are needed to assist healthcare professionals in the early detection of cancer, which typically correlates with more favorable clinical outcomes, or to guide treatment of current cancer patients. Circulating biomarkers are molecules that can be measured in accessible body fluids of individuals, e.g. blood or urine, and whose levels are useful to assist in the diagnosis and/or prognosis and/or prediction of response to treatment. Unfortunately, clinically proven circulating biomarkers are or will be available for only a minority of cancer types, even in cases of advanced metastasis. Examples of widely used biomarkers are the prostate-specific antigen (PSA) for prostate cancer, the carbohydrate antigen 125 (CA125) for ovarian cancer, or the carcinoembryonic antigen (CEA) for colorectal cancer. Even in these cases, the clinical value of these biomarkers for diagnosing cancer is highly debated.

[0750] Emerging cancer biomarkers are circulating nucleic acids directly derived from cancer cells. These nucleic acids can be found for example in circulating free DNA, circulating small-RNA, circulating tumor cells, or extracellular micro-vesicles (including exosomes) containing small-RNA, mRNA and DNA. The minimally invasive technology for detection of such molecular biomarkers without the need for costly or invasive procedures is defined as liquid biopsy. Recently, the definition of liquid biopsy was extended to encompass other macromolecular classes of cancer biomarkers, for example proteins transported within circulating exosomes. It has previously been demonstrated that even circulating metabolites could serve as accurate molecular biomarkers for renal cell carcinoma (RCC), the most common form of kidney cancer (Gatto et al., 2016). In particular, quantitative profiling of glycosaminoglycans (GAGs) in a subjects' blood and/or urine could be scored computationally to derived GAG scores with demonstrated diagnostic valuable for RCC (Gatto et al., 2016).

[0751] In this study, we analyzed data regarding circulating GAG levels in subjects presenting two different cancer types as well as in healthy volunteers. The goal was to identify if alterations from healthy volunteers in the level or chemical composition of specific GAGs were attributable to cancer in general and not significantly affected specifically in a given cancer type, in other words if there exist common patterns of alterations in the GAG profile across different cancers. Furthermore, we sought to analyze transcriptional regulation of GAG metabolism, spanning genes associated with biosynthesis and degradation of GAGs, in a wide spectrum of cancer types by histology. The goal was to identify if common pattern of transcriptional regulation across many different types of cancer may underlie eventual common patterns of alterations in the GAG profile across different cancers.

[0752] Materials and Methods

[0753] Unless otherwise stated, the materials and methods used in this Example correspond to the materials and methods used in Example 1.

[0754] Results

[0755] We analyzed GAG profiles in blood and urine samples from 114 subjects, including 25 healthy volunteers, 100 patients with diagnosis of renal cell carcinoma, and 14 patients with diagnosis of prostate adenocarcinoma. We adopted an ordinary least squares (OLS) to estimate changes in each individual property part of the GAG profiles that were attributable to cancer as opposed to changes attributable specifically to a particular type of cancer (e.g. prostate adenocarcinoma as opposed to renal cell carcinoma). The resulting linear model therefore was:

GAG property.sub.i,j=I.sub.i+.sub.i[Cancer].sub.j+.sub.i[Prostate adenocarcinoma].sub.j+.sub.i

where i indexes a given GAG property in the GAG profile in a certain fluid (blood or urine) and j indexes a given subject, I is the intercept (e.g. the expected value in a healthy subject), is the change attributable to the fact that the subject has cancer and is the change attributable to the fact that the subject has a specific type of cancer, prostate adenocarcinoma, while is the error that accounts for the discrepancy between the actually observed value for the GAG property and the value predicted by the linear modelwhich is minimized in OLS. A statistical test was performed on each estimate for the coefficients in the linear model (I, and ) by computing their t-statistics and calculating the probability p to observe at least a t-value as extreme under the null hypothesis that true coefficient is zero. The null hypothesis was rejected if p was lower than 0.01, after adjusting for multiple testing using the Holm correction (false discovery rate, FDR). This analysis was conducted using base packages in R, version 3.3.1.

[0756] Table D and Table E list the estimates for I (i.e. mean in healthy), (i.e., deviation from mean in cancer) and (i.e. deviation from mean due to cancer-type specific effects) and their respective FDR for each GAG property in the blood or urine, respectively. We focused on those GAG properties that displayed in a particular fluid significant changes attributable to cancer (FDR <0.01) but not to effect specific to a particular cancer types (FDR >0.01).

[0757] In the blood, the following GAG properties were significantly altered in cancer regardless of the cancer type: 0s CS (mass weight fraction, in %); 6s CS (mass weight fraction, in %); 4s/6s CS (ratio of mass weight fractions); Charge CS; Ns HS (mass weight fraction, in %). We observed an increase of 6s CS and Charge CS in the blood of cancer subjects compared to healthy subjects, and a decrease of 0s CS, the 4s/6s CS ratio and Ns HS (FIG. 9A/B/C).

[0758] In the urine, the following GAG properties were significantly altered in cancer regardless of the cancer type: 4s CS (mass weight fraction, in %); 6s/0s CS (ratio of mass weight fractions); 4s/0s (ratio of mass weight fractions); 0s HS (mass weight fraction, in %); Charge HS. We observed an increase of Charge HS in the urine of cancer subjects compared to healthy subjects, and a decrease of 4s CS; 6s/0s CS; 4s/0s CS; 0s HS (FIG. 10A/B/C).

[0759] These results indicate the alteration of specific properties of the GAG profile from normal values in healthy volunteers can be expected in the urine and blood of patients bearing cancer, regardless of the cancer type.

[0760] We sought to design a machine learning classifier to discriminate cancer vs. control subjects using blood and/or urine measurements of the GAG profile. The samples were classified as either cases if they were taken from subjects with a present diagnosis of cancer or controls if they were taken from healthy subjects or subjects with a former diagnosis of cancer but with present no evidence of disease. Therefore, we added 8 subjects to the cohort with a diagnosis of renal cell carcinoma but no evidence of disease at the time of sampling.

[0761] We trained a multilayer perceptron using the open source software weka 3.8.0 (supplied by the University of Waikato, Hamilton, New Zealand) and default parameters on 66% of the cohort samples and tested the accuracy of the model in the remaining 33% of the samples. The results on the ability of the classifier to distinguish cases vs. controls in the training vs. test set for each fluid are shown in Table F. We observed that the classifier had consistently high specificity in the detection of cases ranging 97.5% to 100% in the training set depending on which fluid was used, and this specificity was maintained in the test set with values ranging 81.1% to 95.4%. Importantly, the classifier probed robust to arbitrary choices of the classification threshold, as shown by the fact that classifier was virtually perfect in the training set according to the area under the receiver operating characteristic curve (AUC) with values ranging 0.975 to 1where AUC is a metric that spans an interval from 0 to 1, where 0.5 is a random classifier and 1 is a perfect classifier (FIG. 11). This elevated discriminatory power was observed also in the test set, with AUC values ranging 0.922 to 0.992. These results rule out the possibility that the multilayer perceptron had been overfitted to the training data and suggest that its ability to detect cases could generalize to any independently collected sample.

[0762] We sought to leverage on common cancer-type independent alterations in the GAG profile of cases as opposed to control samples to design blood, urine, or combined GAG scores aimed at discriminating case vs. control subjects. For this purpose, all samples available in our cohort as described in Table F were used. We adopted Lasso penalized logistic regression (Tibshirani, 1996, supra) with leave-one-out cross-validation to select robust GAG properties that were most predictive of the clinical outcome (i.e., case compared to control). Next, we filtered out from this selection those GAG properties for which we observed that the difference in mean between case vs. control had an estimated highest density interval (HDI) which included the value 0. The HDI for each GAG property was calculated using Bayesian estimation under the following assumptions: for each property, the GAG property values were assumed to be sampled from a t-distribution of unknown and to be estimated normality (i.e. degrees of freedom); high uncertainty on the prior distributions; the marginal distribution was well approximated by a Markov chain Monte Carlo sampling with no thinning and chain length equal to 100'000. The estimation was performed using BEST R-package (the above assumptions were reflected by the default parameters). Each GAG score was then subsequently designed as a ratio, where the numerator was the sum of the filtered properties associated with case and the denominator was the sum of the filtered properties associated with control. Each term was normalized using the regression coefficients from the logistic regression. The following GAG scores were designed:

[00017] $Blood .Math. .Math. score = \frac{10 [6 .Math. s .Math. .Math. CS] + 20 [\frac{6 .Math. s}{0 .Math. s} .Math. CS]}{\frac{4}{100} [\frac{4 .Math. s}{6 .Math. s} .Math. CS] + [0 .Math. s .Math. .Math. CS]}$ $Urine .Math. .Math. score = 2 .Math. \frac{[Ns .Math. .Math. 2 .Math. s .Math. .Math. HS]}{.Math. 4 .Math. s .Math. .Math. CS .Math.}$ $Combined .Math. .Math. score = \frac{(Blood .Math. .Math. score + Urine .Math. .Math. score)}{2}$

where terms in brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG. We subsequently calculated the three scores for each sample and observed that case samples have recurrently elevated scores with respect to control samples (FIGS. 12, 13 and 14). The difference in GAG scores between case and controls probed able to discriminate subjects with very high accuracy in that ROC curves generated for each GAG score had an AUC equal to 0.961 for blood, 0.942 for urine, and 0.998 for combined (FIG. 15). Overall, these results indicate that scores can be designed from blood and/or urine measurements of specific GAG properties to detect cancer regardless of the cancer type.

[0763] We also retrieved gene expression data for 19 cancer subtypes encompassing 17 cancer types by histology: bile duct cancer, bladder cancer, blood cancer, brain cancer, breast cancer, colorectal cancer, head and neck cancer, liver cancer, lung cancer, oesophageal cancer, ovarian cancer, pancreatic cancer, prostate cancer, skin cancer, stomach cancer, thyroid cancer, and uterine cancer. Gene expression data in the form of RNAseq-generated read count tables was retrieved for 16 of these subtypes from the Cancer Genome Atlas (TCGA) project (ttps://gdc-portal.nci.nih.gov/). Gene expression data in the form of microarray-generated signal intensity tables was retrieved for the remaining 3 subtypes from three public datasets deposited at the Gene Expression Omnibus (GEO) (Raskin et al., 2013, Mok et al., 2009, Iqbal et al., 2010). We also retrieved TCGA data for renal cell carcinoma (RCC) and three subtypes of RCC: chromophobe RCC, clear cell RCC, and papillary RCC. Both primary tumour tissue samples and matched normal tissues samples were retrieved for each cancer subtype. A summary of the data retrieved for this analysis in reported in Table G.

[0764] The gene expression data was pre-processed as previously described (Ritchie et al., 2015) using the limma R-package (Smyth, 2004). Differential gene expression analysis to assess changes in transcriptional regulation between tumor and normal samples was performed separately for each cancer subtype using the limma R-package in the case of microarray data and using the voom R-package (Law et al., 2014) in the case of RNA-seq data, as previously described (Ritchie et al., 2015). For a given gene, changes in transcriptional regulation in either direction, e.g. up-regulation or down-regulation, from normal tissues were considered statistically significant if we observed a false discovery rate q<0.001 and a minimum absolute fold-change in expression level >50%.

[0765] Next, we focused on 60 genes associated with GAG metabolism. A complete list of genes is shown in Table H. We observed a cancer subtype-specific profile of differential transcriptional regulation in comparison to normal tissues (FIG. 16 and Table K). On average, 30 of 60 (50%) GAG genes were differentially regulated in each cancer type (12 down-regulated and 18 up-regulated). At least 2 GAG genes were differentially regulated in all cancer subtypes analyzed. The maximum number of differentially regulated GAG genes was 39, which was observed in cholangiocarcinoma. None of the analyzed cancer subtypes displayed a profile of transcriptional regulation of GAG genes identical to any of the other subtypes. These results suggest that each cancer subtype can affect the level and chemical composition of GAGs in a cancer subtype-specific fashion, which may reflect the cancer-type specific effects observed in GAG profile alterations noted above. Strikingly, we observed GAG genes that displayed the same pattern of transcriptional regulation virtually in all cancer types examined: CHPF2 was up-regulated in 17 of 22 cancer types, CHPF and B4GALT7 in 16 of 22 types and B3GAT3 in 15 of 22 types. None of these was found to be significantly down-regulated in the remaining types. EXTL1 and HPSE2 were significantly down-regulated in 11 and 12 of 22 types respectively. None of these was found to be significantly up-regulated in the remaining types. These patterns underlie regulatory programs vastly conserved across different cancer types.

[0766] B4GALT7 and B3GAT3 encode for enzymes required to synthesize the linkage region of any GAGs, while CHPF and CHPF2 encode for the primary enzymes responsible for the polymerization of CS by adding non-sulfated CS monomers to the growing chain. Their up-regulation in most cancers can be expected to alter the level and composition of the emerging GAGs by increasing the degree of non-sulfated CS, resulting in the common patterns here observed in the urine of two different cancer types, where the sulfated fraction was significantly reduced compared to the unsulfated fraction (FIG. 10B/C). The opposite pattern in the blood can be ascribed as the physiological consequence of GAG filtration from the blood stream operated by the kidneys, which results in an enrichment of sulfated CS species in the blood (FIG. 9B/C). On the contrary, repression of EXTL1, which participates in the polymerization of HS by adding non-sulfated HS monomers to the linkage region, may decrease the degree of non-sulfated HS, resulting the common pattern here observed in the urine of two different cancer types, where the sulfated fraction was significantly increased compared to the unsulfated fraction (FIG. 10A). Once again, the physiological GAG filtration operated by the kidneys will result in the opposite pattern in the blood, where one sulfated form of HS was significantly decreased (FIG. 9A). Repression of HPSE2, known to inhibit of HPSE, may elicit the heparanase activity of HPSE, which is aimed at degrading mature HS polymers thus exacerbating the reduction of nonsulfated HS normally observed in healthy individuals.

[0767] These observations suggest that common patterns in the regulation of GAG metabolism across cancers are linked to common alterations observed in the urine and blood GAG profile in different cancer types. This strengthens the conclusions that specific GAG properties are expected to be altered in body fluids compared to healthy conditions as a consequence of cancer, regardless of the cancer type.

TABLE-US-00004 TABLE D Estimates for mean values of each GAG property in the blood of healthy volunteers and their change in presence of cancer and of cancer-type specific effects. FDR for each estimate are reported with FDR < 0.01 considered statistically significant (bold underline). GAG properties in the blood that displayed significant changes attributable to cancer but not to effect specific to a particular cancer types are highlighted in underline. FDR Deviation deviation from mean from mean due to FDR due to Deviation cancer-type FDR deviation cancer-type Mean in from mean specific mean in from mean specific GAG property healthy in cancer effects health in cancer effects 0sCSblood 81.87 9.88 3.24 9.61E70 5.76E03 1.00E+00 2 s.sub.CS.sub.blood 0.03 0.08 0.01 1.00E+00 9.02E01 1.00E+00 6sCSblood 0.36 2.79 1.81 1.00E+00 1.94E05 2.87E01 4 s.sub.CS.sub.blood 17.67 6.70 4.93 2.61E14 9.62E02 1.00E+00 2 s6 s.sub.CS.sub.blood 0.03 0.07 0.04 1.00E+00 4.91E01 1.00E+00 2 s4 s.sub.CS.sub.blood 0.01 0.04 0.11 1.00E+00 1.00E+00 5.09E02 4 s6 s.sub.CS.sub.blood 0.03 0.11 0.01 1.00E+00 1.84E01 1.00E+00 Tris.sub.CS.sub.blood 0.01 0.04 0.04 1.00E+00 1.00E+00 1.00E+00 4s/6sCSblood 68.77 56.37 15.16 3.41E41 1.51E28 8.80E02 6 s/0 s.sub.CS.sub.blood 0.00 0.05 0.03 1.00E+00 1.64E01 1.00E+00 4 s/0 s.sub.CS.sub.blood 0.22 0.17 0.04 1.69E03 1.32E01 1.00E+00 ChargeCSblood 0.18 0.10 0.04 2.84E11 5.81E03 1.00E+00 Tot.sub.HA.sub.blood 0.12 0.04 0.06 5.78E10 8.45E01 3.52E01 Tot.sub.CS.sub.blood 5.37 0.18 1.20 1.21E13 1.00E+00 1.00E+00 Tris.sub.HS.sub.blood 1.18 1.00 1.03 3.82E01 1.00E+00 1.00E+00 Ns6 s.sub.HS.sub.blood 1.69 0.79 1.28 1.11E04 1.00E+00 3.52E01 Ns2 s.sub.HS.sub.blood 1.69 0.98 0.12 1.69E03 9.02E01 1.00E+00 NsHSblood 16.44 10.18 5.14 7.87E25 6.72E10 1.62E01 2 s6 s.sub.HS.sub.blood 1.99 0.28 1.51 9.40E05 1.00E+00 3.31E01 6 s.sub.HS.sub.blood 6.87 0.24 9.10 1.68E05 1.00E+00 9.76E05 2 s.sub.HS.sub.blood 5.35 1.21 6.30 1.71E05 1.00E+00 9.87E04 0 s.sub.HS.sub.blood 64.79 8.55 16.63 1.98E45 3.31E01 4.15E03 Charge.sub.HS.sub.blood 0.43 0.04 0.12 1.70E17 1.00E+00 1.00E+00 Tot.sub.HS.sub.blood 0.05 0.08 0.06 3.17E01 2.09E02 9.02E01

TABLE-US-00005 TABLE E Estimates for mean values of each GAG property in the urine of healthy volunteers and their change in presence of cancer and of cancer-type specific effects. FDR for each estimate are reported with FDR < 0.01 considered statistically significant (bold underline). GAG properties in the urine that displayed significant changes attributable to cancer but not to effect specific to a particular cancer types are highlighted in underline. FDR Deviation deviation from mean from mean due to FDR due to Deviation cancer-type FDR deviation cancer-type Mean in from mean specific mean in from mean specific GAG property healthy in cancer effects health in cancer effects 0 s.sub.CS.sub.urine 35.01 23.38 17.87 5.81E18 4.16E06 5.81E03 2 s.sub.CS.sub.urine 0.10 0.00 0.02 7.32E05 1.00E+00 1.00E+00 6 s.sub.CS.sub.urine 21.65 7.94 12.16 3.57E20 1.29E02 2.81E04 4sCSurine 40.91 14.72 4.28 4.69E34 1.33E07 1.00E+00 2 s6 s.sub.CS.sub.urine 0.61 0.16 0.43 1.07E13 1.00E+00 2.39E03 2 s4 s.sub.CS.sub.urine 0.46 0.13 0.30 4.27E16 4.40E01 5.91E04 4 s6 s.sub.CS.sub.urine 1.18 0.37 0.59 4.02E16 3.01E01 2.54E02 Tris.sub.CS.sub.urine 0.08 0.05 0.09 1.10E03 1.00E+00 1.60E01 4 s/6 s.sub.CS.sub.urine 2.02 0.43 1.24 8.83E16 1.00E+00 2.41E03 6s/0sCSurine 0.76 0.47 0.41 1.06E12 3.79E03 7.46E02 4s/0sCSurine 1.43 0.92 0.30 1.19E13 9.33E04 1.00E+00 Charge.sub.CS.sub.urine 0.65 0.24 0.21 7.16E31 3.75E06 6.25E04 Tot.sub.HA.sub.urine 0.07 0.13 0.05 7.90E01 3.43E01 1.00E+00 Tot.sub.CS.sub.urine 0.98 3.85 1.62 1.00E+00 5.63E02 1.00E+00 Tris.sub.HS.sub.urine 0.44 0.69 1.02 9.50E01 8.53E01 2.81E01 Ns6 s.sub.HS.sub.urine 1.63 1.07 0.70 5.91E04 8.53E01 1.00E+00 Ns2 s.sub.HS.sub.urine 1.49 1.71 1.74 1.06E05 1.40E03 7.97E03 Ns.sub.HS.sub.urine 12.67 6.32 5.49 4.02E12 5.95E02 4.40E01 2 s6 s.sub.HS.sub.urine 1.18 0.37 0.24 2.39E03 1.00E+00 1.00E+00 6 s.sub.HS.sub.urine 7.28 4.10 2.14 2.53E08 1.93E01 1.00E+00 2 s.sub.HS.sub.urine 3.81 1.88 3.14 2.78E06 8.64E01 1.37E01 0sHSurine 71.49 16.13 6.87 3.08E35 8.32E04 1.00E+00 ChargeHSurine 0.34 0.21 0.02 8.72E16 2.79E04 1.00E+00 Tot.sub.HS.sub.urine 0.15 0.66 0.41 1.00E+00 4.45E01 1.00E+00

TABLE-US-00006 TABLE F Metrics of accuracy for a multilayer perceptron classifier trained on 66% of the samples and tested on the remaining 33% of the samples. Three classifier were trained based on either only the blood GAG profile, or only on the urine GAG profile, or on both blood and urine GAG profiles. Samples were classified as either cases, if they were taken from subjects with a present diagnosis of cancer, or controls, if they were taken from healthy subjects or subjects with a former diagnosis of cancer but with present no evidence of disease. Samples Samples in train set in test set AUC in Sensitivity Specificity AUC in Sensitivity Specificity Fluid (N cases) (N cases) test set in train set in train set test set in train set in train set Plasma 87 (69) 47 (37) 1 100% 100% 0.922 .sup.83% 95.4% Urine 40 (22) 22 (12) 0.965 97.5% 98% 0.992 81.8% 84.8% Both plasma 39 (21) 22 (12) 1 100% 100% 0.975 77.3% 81.1% and urine

TABLE-US-00007 TABLE G Number of samples analyzed for each cancer subtype and their source. N tumor N normal Cancer subtype Abbreviation Cancer type samples samples Source Clear cell renal cell carcinoma KIRC Renal cell 532 72 TCGA carcinoma Papillary renal cell carcinoma KIRP Renal cell 289 32 TCGA carcinoma Chromophobe renal cell KICH Renal cell 66 25 TCGA carcinoma carcinoma Thyroid carcinoma THCA Thyroid cancer 505 59 TCGA Prostate adenocarcinoma PRAD Prostate cancer 496 52 TCGA Colon adenocarcinoma COAD Colorectal 282 41 TCGA cancer Rectum adenocarcinoma READ Colorectal 94 10 TCGA cancer Lung squamous cell LUSC Lung cancer 496 51 TCGA carcinoma Uterine corpus endometrial UCEC Uterine cancer 174 24 TCGA carcinoma Breast invasive carcinoma BRCA Breast cancer 1090 114 TCGA Pancreatic adenocarcinoma PAAD Pancreatic 178 4 TCGA cancer Lung adenocarcinoma LUAD Lung cancer 514 59 TCGA Bladder carcinoma BLCA Bladder cancer 407 19 TCGA Liver hepatocellular carcinoma LIHC Liver cancer 362 50 TCGA Cholangiocarcinoma CHOL Bile duct cancer 36 9 TCGA Stomach adenocarcinoma STAD Stomach cancer 415 35 TCGA Oesophageal carcinoma ESCA Oesophageal 184 11 TCGA cancer Head and neck squamous cell HNSC Head and neck 519 44 TCGA carcinoma cancer Glioblastoma multiforme GBM Brain cancer 153 5 TCGA Peripheral T-cell lymphoma PTCL Blood cancer 131 10 GEO (GSE19069) Serous ovarian OVSA Ovarian cancer 53 10 GEO adenocarcinoma (GSE18520) Malignant melanoma MM Skin cancer 46 16 GEO (GSE15605)

TABLE-US-00008 TABLE H List of genes associated with GAG metabolism. GAG metabolic Gene symbol process Gene description UST CS synthesis uronyl-2-sulfotransferase [Source: HGNC Symbol; Acc: 17223] CHST14 CS synthesis carbohydrate (N-acetylgalactosamine 4-0) sulfotransferase 14 [Source: HGNC Symbol; Acc: 24464] CHST13 CS synthesis carbohydrate (chondroitin 4) sulfotransferase 13 [Source: HGNC Symbol; Acc: 21755] CHSY1 CS synthesis chondroitin sulfate synthase 1 [Source: HGNC Symbol; Acc: 17198] DSE CS synthesis dermatan sulfate epimerase [Source: HGNC Symbol; Acc: 21144] CHSY3 CS synthesis chondroitin sulfate synthase 3 [Source: HGNC Symbol; Acc: 24293] CHST11 CS synthesis carbohydrate (chondroitin 4) sulfotransferase 11 [Source: HGNC Symbol; Acc: 17422] CHST15 CS synthesis carbohydrate (N-acetylgalactosamine 4-sulfate 6-O) sulfotransferase 15 [Source: HGNC Symbol; Acc: 18137] CHPF2 CS synthesis chondroitin polymerizing factor 2 [Source: HGNC Symbol; Acc: 29270] CSGALNACT2 CS synthesis chondroitin sulfate N-acetylgalactosaminyltransferase 2 [Source: HGNC Symbol; Acc: 24292] CHST12 CS synthesis carbohydrate (chondroitin 4) sulfotransferase 12 [Source: HGNC Symbol; Acc: 17423] CSGALNACT1 CS synthesis chondroitin sulfate N-acetylgalactosaminyltransferase 1 [Source: HGNC Symbol; Acc: 24290] CHST7 CS synthesis carbohydrate (N-acetylglucosamine 6-O) sulfotransferase 7 [Source: HGNC Symbol; Acc: 13817] CHPF CS synthesis chondroitin polymerizing factor [Source: HGNC Symbol; Acc: 24291] CHST3 CS synthesis carbohydrate (chondroitin 6) sulfotransferase 3 [Source: HGNC Symbol; Acc: 1971] HPSE GAG degradation heparanase [Source: HGNC Symbol; Acc: 5164] HGSNAT GAG degradation heparan-alpha-glucosaminide N-acetyltransferase [Source: HGNC Symbol; Acc: 26527] HYAL4 GAG degradation hyaluronoglucosaminidase 4 [Source: HGNC Symbol; Acc: 5323] GALNS GAG degradation galactosamine (N-acetyl)-6-sulfate sulfatase [Source: HGNC Symbol; Acc: 4122] GLB1 GAG degradation galactosidase, beta 1 [Source: HGNC Symbol; Acc: 4298] GNS GAG degradation glucosamine (N-acetyl)-6-sulfatase [Source: HGNC Symbol; Acc: 4422] GUSB GAG degradation glucuronidase, beta [Source: HGNC Symbol; Acc: 4696] HEXA GAG degradation hexosaminidase A (alpha polypeptide) [Source: HGNC Symbol; Acc: 4878] HEXB GAG degradation hexosaminidase B (beta polypeptide) [Source: HGNC Symbol; Acc: 4879] HYAL1 GAG degradation hyaluronoglucosaminidase 1 [Source: HGNC Symbol; Acc: 5320] IDS GAG degradation iduronate 2-sulfatase [Source: HGNC Symbol; Acc: 5389] IDUA GAG degradation iduronidase, alpha-L- [Source: HGNC Symbol; Acc: 5391] ARSB GAG degradation arylsulfatase B [Source: HGNC Symbol; Acc: 714] NAGLU GAG degradation N-acetylglucosaminidase, alpha [Source: HGNC Symbol; Acc: 7632] HPSE2 GAG degradation heparanase 2 [Source: HGNC Symbol; Acc: 18374] SGSH GAG degradation N-sulfoglucosamine sulfohydrolase [Source: HGNC Symbol; Acc: 10818] SPAM1 GAG degradation sperm adhesion molecule 1 (PH-20 hyaluronidase, zona pellucida binding) [Source: HGNC Symbol; Acc: 11217] HYAL3 GAG degradation hyaluronoglucosaminidase 3 [Source: HGNC Symbol; Acc: 5322] HYAL2 GAG degradation hyaluronoglucosaminidase 2 [Source: HGNC Symbol; Acc: 5321] HS3ST3B1 GAG degradation heparan sulfate (glucosamine) 3-O-sulfotransferase 3B1 [Source: HGNC Symbol; Acc: 5198] HS3ST3B1 GAG degradation heparan sulfate (glucosamine) 3-O-sulfotransferase 3B1 [Source: HGNC Symbol; Acc: 5198] HS3ST3A1 GAG degradation heparan sulfate (glucosamine) 3-O-sulfotransferase 3A1 [Source: HGNC Symbol; Acc: 5196] HS3ST3A1 GAG degradation heparan sulfate (glucosamine) 3-O-sulfotransferase 3A1 [Source: HGNC Symbol; Acc: 5196] B4GALT7 GAG synthesis xylosylprotein beta 1,4-galactosyltransferase, polypeptide 7 [Source: HGNC Symbol; Acc: 930] B3GALT6 GAG synthesis UDP-Gal: betaGal beta 1,3-galactosyltransferase polypeptide 6 [Source: HGNC Symbol; Acc: 17978] B3GAT2 GAG synthesis beta-1,3-glucuronyltransferase 2 (glucuronosyltransferase S) [Source: HGNC Symbol; Acc: 922] B3GAT3 GAG synthesis beta-1,3-glucuronyltransferase 3 (glucuronosyltransferase I) [Source: HGNC Symbol; Acc: 923] B3GAT1 GAG synthesis beta-1,3-glucuronyltransferase 1 (glucuronosyltransferase P) [Source: HGNC Symbol; Acc: 921] XYLT1 GAG synthesis xylosyltransferase I [Source: HGNC Symbol; Acc: 15516] XYLT2 GAG synthesis xylosyltransferase II [Source: HGNC Symbol; Acc: 15517] EXT1 HS synthesis exostosin glycosyltransferase 1 [Source: HGNC Symbol; Acc: 3512] EXT2 HS synthesis exostosin glycosyltransferase 2 [Source: HGNC Symbol; Acc: 3513] EXTL1 HS synthesis exostosin-like glycosyltransferase 1 [Source: HGNC Symbol; Acc: 3515] EXTL2 HS synthesis exostosin-like glycosyltransferase 2 [Source: HGNC Symbol; Acc: 3516] EXTL3 HS synthesis exostosin-like glycosyltransferase 3 [Source: HGNC Symbol; Acc: 3518] HS3ST5 HS synthesis heparan sulfate (glucosamine) 3-O-sulfotransferase 5 [Source: HGNC Symbol; Acc: 19419] GLCE HS synthesis glucuronic acid epimerase [Source: HGNC Symbol; Acc: 17855] HS6ST3 HS synthesis heparan sulfate 6-O-sulfotransferase 3 [Source: HGNC Symbol; Acc: 19134] NDST1 HS synthesis N-deacetylase/N-sulfotransferase (heparan glucosaminyl) 1 [Source: HGNC Symbol; Acc: 7680] NDST4 HS synthesis N-deacetylase/N-sulfotransferase (heparan glucosaminyl) 4 [Source: HGNC Symbol; Acc: 20779] HS6ST2 HS synthesis heparan sulfate 6-O-sulfotransferase 2 [Source: HGNC Symbol; Acc: 19133] NDST3 HS synthesis N-deacetylase/N-sulfotransferase (heparan glucosaminyl) 3 [Source: HGNC Symbol; Acc: 7682] HS6ST1 HS synthesis heparan sulfate 6-O-sulfotransferase 1 [Source: HGNC Symbol; Acc: 5201] HS2ST1 HS synthesis heparan sulfate 2-O-sulfotransferase 1 [Source: HGNC Symbol; Acc: 5193] HS3ST2 HS synthesis heparan sulfate (glucosamine) 3-O-sulfotransferase 2 [Source: HGNC Symbol; Acc: 5195] HS3ST1 HS synthesis heparan sulfate (glucosamine) 3-O-sulfotransferase 1 [Source: HGNC Symbol; Acc: 5194] NDST2 HS Synthesis N-deacetylase and N-sulfotransferase 2 (e.g. Entrez [Source: HGNC] ID: 8509)

TABLE-US-00009 TABLE K Regulation of GAG genes in different cancer types vs. matched normal tissue of origin. Entries with 1 indicate that the gene in that row was found to be up-regulated in the cancer type in that column; 1 indicate that the gene in that row was found to be down-regulated in the cancer type in that column; 0 indicate that the gene in that row was not found to be regulated in the cancer type in that column; NA indicate unavailable information. Genes are identified by their official gene symbol or, in not available, by their entrez gene ID. Key as in Table G. Gene BLCA BRCA COAD GBM HNSC LUSC LUAD PAAD READ UCEC CHOL B3GALT6 1 1 1 0 1 0 0 0 1 0 1 XYLT1 0 1 1 0 1 0 0 0 0 1 NA XYLT2 1 1 1 0 1 1 1 0 0 0 1 B3GAT1 0 1 1 0 0 1 1 0 NA 0 0 B4GALT7 1 1 1 1 1 1 1 0 1 1 1 B3GAT2 1 1 0 0 NA 0 0 NA NA NA NA B3GAT3 1 1 1 0 1 1 1 0 1 1 1 DSE 0 1 0 1 1 0 0 0 0 1 1 CHPF 1 1 1 1 1 1 1 1 1 1 1 CHSY3 0 0 1 0 1 1 0 0 1 1 0 CHST14 1 0 1 1 1 1 0 0 0 1 1 CHSY1 0 0 1 1 1 0 1 0 1 1 1 CHST11 1 1 0 1 1 1 0 0 0 0 NA CHPF2 1 1 0 1 1 1 1 0 1 1 1 CHST15 0 1 1 0 1 0 1 0 1 0 1 CHST7 0 1 1 1 1 1 1 0 0 1 0 UST 1 1 1 1 1 0 1 0 1 1 0 CHST13 NA 0 1 NA NA 1 0 NA 0 NA 1 CSGALNACT2 0 0 1 0 1 1 0 0 0 1 1 CSGALNACT1 1 1 1 0 0 1 1 0 0 1 1 CHST3 0 1 0 1 1 1 0 0 0 1 1 CHST12 1 0 0 1 1 0 0 0 0 0 1 HS3ST5 1 0 1 1 1 1 1 NA NA NA NA EXT2 0 0 0 1 1 1 0 0 0 0 1 EXT1 0 0 1 1 1 1 0 0 1 0 1 NDST3 NA 0 0 1 NA NA NA NA NA NA NA NDST1 0 1 1 0 0 1 1 0 0 0 1 GLCE 1 0 1 0 0 0 0 0 0 0 1 HS2ST1 1 0 1 1 1 0 0 0 1 0 1 HS6ST2 1 1 1 0 0 1 1 NA 1 1 NA HS3ST1 0 1 1 1 1 1 1 1 0 0 1 HS3ST2 0 0 0 1 1 1 0 0 1 0 NA HS6ST1 1 1 0 0 0 1 0 0 0 1 1 8509 0 0 0 0 0 0 0 0 0 0 1 NDST4 NA NA NA NA NA NA NA NA NA NA NA HS6ST3 0 1 1 1 NA 0 0 NA NA 1 NA EXTL1 1 0 1 1 1 1 0 NA NA 1 NA EXTL2 0 1 1 1 1 0 0 0 0 1 1 EXTL3 0 1 1 0 1 0 0 0 0 0 1 HS3ST3B1 NA 0 1 NA 0 0 0 NA NA 0 1 HS3ST3A1 0 1 0 NA 0 1 1 0 0 0 0 HPSE 1 1 1 NA 0 1 1 0 1 0 NA HPSE2 1 1 1 NA 1 1 1 NA 1 1 NA GLB1 1 1 0 1 1 1 0 0 0 1 1 GUSB 1 0 0 1 1 0 0 0 0 0 0 HYAL3 1 1 1 0 0 1 1 0 1 1 0 GNS 0 1 0 1 1 1 0 0 0 0 1 HYAL4 NA NA NA NA 1 NA NA NA NA NA NA HYAL1 1 1 0 NA 0 1 1 0 0 1 1 HYAL2 0 0 1 1 0 1 1 0 1 0 1 SPAM1 NA NA NA NA NA NA NA NA NA NA NA HGSNAT 0 0 0 0 0 0 0 0 0 1 1 GALNS 1 1 0 1 1 1 0 0 0 1 1 IDUA 1 0 0 1 0 1 0 0 1 0 1 SGSH 1 0 0 1 0 0 0 0 0 0 1 HEXB 0 0 0 1 1 1 0 0 0 0 1 NAGLU 1 0 0 1 1 0 0 0 0 0 0 HEXA 1 0 0 1 1 0 0 0 0 0 1 IDS 0 1 0 1 0 1 0 0 0 1 1 ARSB 0 0 0 1 1 1 0 0 0 0 1 Gene ESCA KICH KIRC KIRP LIHC STAD THCA PRAD MM OVSA PTCL B3GALT6 1 0 0 0 1 1 0 1 1 1 0 XYLT1 0 1 0 0 0 0 1 0 0 0 0 XYLT2 0 0 0 1 1 0 0 0 0 0 0 B3GAT1 1 1 1 0 1 1 1 1 0 0 0 B4GALT7 1 1 1 1 1 1 0 0 0 0 0 B3GAT2 0 NA 1 NA NA 0 1 0 NA NA NA B3GAT3 1 1 0 1 1 1 0 1 0 0 0 DSE 1 0 0 0 1 0 0 1 0 0 0 CHPF 1 0 1 1 0 1 0 1 0 0 0 CHSY3 1 0 1 0 0 1 0 0 1 0 1 CHST14 1 0 1 1 1 0 1 0 1 1 0 CHSY1 1 0 1 0 0 1 0 0 1 1 0 CHST11 1 0 1 1 1 1 1 1 1 0 0 CHPF2 1 1 1 1 1 1 1 0 1 0 0 CHST15 1 1 1 0 1 0 0 1 0 0 1 CHST7 1 1 1 0 1 0 1 0 0 0 0 UST 0 1 1 1 1 1 1 1 0 0 0 CHST13 NA 1 1 1 1 1 NA NA 0 0 0 CSGALNACT2 1 1 1 0 0 1 1 0 0 0 0 CSGALNACT1 0 1 0 1 1 0 1 1 0 0 0 CHST3 0 0 1 0 1 0 0 1 0 0 0 CHST12 1 0 1 1 1 0 0 0 0 0 0 HS3ST5 NA 1 1 NA NA 0 NA 1 NA NA NA EXT2 1 0 0 0 1 1 0 0 0 0 0 EXT1 1 0 0 0 1 1 1 1 0 0 1 NDST3 1 1 1 1 1 0 1 NA NA NA NA NDST1 0 1 0 1 1 0 0 0 0 1 0 GLCE 1 1 1 1 1 1 0 1 0 0 0 HS2ST1 1 1 0 0 1 1 0 0 0 0 0 HS6ST2 1 1 1 1 NA 1 1 0 0 0 0 HS3ST1 1 0 1 1 0 1 0 0 0 0 0 HS3ST2 0 NA 1 NA 1 1 0 1 0 1 0 HS6ST1 1 1 1 1 1 0 1 0 0 0 0 8509 0 0 0 0 1 0 0 0 NA NA NA NDST4 NA NA NA NA NA 1 NA 0 NA NA NA HS6ST3 1 1 1 1 NA 1 1 1 0 0 0 EXTL1 NA 1 1 1 NA 1 0 1 0 0 0 EXTL2 0 1 1 0 1 1 0 0 0 0 0 EXTL3 1 0 0 0 1 0 0 0 0 0 0 HS3ST3B1 0 1 1 1 1 0 1 1 0 0 0 HS3STSA1 1 1 1 1 1 0 1 1 0 0 1 HPSE 1 0 0 1 0 1 1 1 0 0 0 HPSE2 1 NA 1 NA NA 1 NA 1 0 0 0 GLB1 1 1 1 0 1 1 0 0 NA NA NA GUSB 1 0 1 1 0 1 0 1 0 0 0 HYAL3 1 0 1 0 1 0 0 1 0 0 0 GNS 0 0 1 0 1 1 0 0 1 0 0 HYAL4 NA 0 1 1 NA NA NA NA 0 0 0 HYAL1 0 1 1 1 0 0 0 1 0 0 0 HYAL2 0 1 0 1 1 0 1 0 0 0 0 SPAM1 NA NA NA NA NA NA NA NA NA NA NA HGSNAT 0 0 0 0 0 0 0 0 0 0 0 GALNS 1 0 1 1 0 1 0 0 0 0 0 IDUA 0 0 1 1 1 0 0 1 0 0 0 SGSH 0 0 0 1 1 0 0 1 0 0 0 HEXB 1 0 0 1 1 1 0 0 0 0 0 NAGLU 0 0 0 0 0 0 0 0 0 0 0 HEXA 1 1 0 0 1 1 0 0 1 0 0 IDS 1 1 0 0 0 0 0 1 0 0 0 ARSB 0 1 0 0 1 1 0 0 1 0 0

REFERENCES

[0768] 1. F. Gatto et al., Glycosaminoglycan Profiling in Patients' Plasma and Urine Predicts the Occurrence of Metastatic Clear Cell Renal Cell Carcinoma. Cell Rep 15, 1822-1836 (2016). [0769] 2. L. Raskin et al., Transcriptome profiling identifies HMGA2 as a biomarker of melanoma progression and prognosis. J Invest Dermatol 133, 2585-2592 (2013). [0770] 3. S. C. Mok et al., A gene signature predictive for outcome in advanced ovarian cancer identifies a survival factor: microfibril-associated glycoprotein 2. Cancer Cell 16, 521-532 (2009). [0771] 4. J. Iqbal et al., Molecular signatures to improve diagnosis in peripheral T-cell lymphoma and prognostication in angioimmunoblastic T-cell lymphoma. Blood 115, 1026-1036 (2010). [0772] 5. M. E. Ritchie et al., limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47 (2015). [0773] 6. G. K. Smyth, Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3, Article3 (2004). [0774] 7. C. W. Law, Y. Chen, W. Shi, G. K. Smyth, Voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biol 15, R29 (2014).

EXAMPLE 3

[0775] In this study we measured circulating GAG levels in subjects presenting with melanoma (M).

[0776] Material and Methods

[0777] Plasma Sample Collection

[0778] Blood samples were obtained from 14 healthy volunteers (H) and 16 patients with uveal (N=12) or skin melanoma (N=4). All patients had evidence of disease at the time of sampling.

[0779] Whole blood was collected in EDTA tubes and centrifuged first at 1,000 g for 10 minutes at 4 C. within 30 minutes from collection and then the supernatant was centrifuged again at 10,000 g for 10 minutes at 4 C. The supernatant is platelet-poor plasma was frozen at 80 C. until they were shipped for analysis in dry ice.

[0780] Platelet-poor plasma GAGs correspond to blood GAGs. For the purpose of this study, platelet-poor plasma is therefore also interchangeably referred to as blood.

[0781] Glycosaminoglycan Profile Determination

[0782] Sample preparation including extraction and purification steps were performed as previously described by Volpi and Maccari 2005.sup.1 and 2 and Coppa et al., 2011, while sample GAGs separation and quantification were performed as described in Volpi et al., 2014; Volpi and Linhardt, 2010.

[0783] Briefly, to extract the GAGs, 500 l of sample was lyophilized, reconstituted with 1 ml of a 20-mM TRIS-CI buffer pH 7.4 and treated with protease (Proteinase K from Tritirachium album [E.C. 3.4.21.64], >500 units ml_1 from Sigma-Aldrich) at 60 C. for 12 h. After boiling for 10 min, centrifugation and filtration on 0.45-m filters, the filtrate was lyophilized. The powder was dissolved in 1 ml of distilled water by prolonged mixing. After centrifugation at 5000 g for 15 min, 0.2 ml of 20% trichloro-acetic acid was added to the supernatant. After 2 h at 4 C., the mixture was centrifuged at 5000 g for 15 min, and the supernatant was recovered and lyophilized. After solubilization in 0.4 ml of bidistilled water and centrifugation at 10 000 g for 10 min, the supernatant was collected and further analysed. To purify the GAGs, after reconstitution with 500 l 10 mM NaCl, sample GAGs were further purified on anion-exchange resin (QAE Sephadex A-25). After centrifugation at 10,000g for 5 min, the supernatant was applied to a column (1 cm4 cm) packed with about 3 ml of resin previously equilibrated with 10 mM NaCl. After washing the resin with 20 ml of 10 mM NaCl, 10 ml of 2.5 M NaCl were added. Fifty milliliters of ethanol were added to the eluate (10 ml) and stored at 20 C. for 24 h. After centrifugation at 5000g for 15 min, the pellet was dried at 60 C. for 12 h. After reconstitution with 80 l of 50 mM ammonium acetate pH 8.0, the material was treated with 20 l of chondroitinase ABC at 37 C. for 12 h. To separate and quantify the GAGs, after boiling for 5 min, the samples were injected in a capillary electrophoresis instrument equipped with laser-induced fluorescence detector (CE-LIF).

[0784] Nineteen independent GAG properties were measured in each sample (blood): CS concentration (in g/mL), HS concentration (in g/mL), HA concentration (in g/mL), and mass fractions of disaccharide composition for both CS and HS. In addition, five dependent GAG properties were calculated from these measurements: the CS and HS charge; and the following CS ratio of mass fractions: 4s/6s, 6s/0s, and 4s/0s. The 24 GAG properties may be collectively referred to as GAG profile.

[0785] Biomarker Score Design

[0786] The statistical significance of changes for each GAG property between the two groups (melanoma vs. healthy) was assessed by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation under the following assumptions: GAG property values are sampled from a t-distribution of unknown and to be estimated normality (i.e. degrees of freedom); high uncertainty on the prior distributions; the marginal distribution is well approximated by a Markov chain Monte Carlo sampling with no thinning and chain length equal to 100'000. The estimation was performed using BEST R-package (the above assumptions are reflected by the default parameters). Bayesian estimation was preferred over the widely used t-test since it provides a robust and reliable estimation of mean difference even under uncertainty of the underlying score distribution for the two groups (that is the case when the number of samples is limited).

[0787] To verify whether changes in the GAG profile in melanoma patients could be condensed into GAG scores, that can be used to differentiate melanoma patients from healthy subjects, and therefore would be suitable as a cancer diagnostic, we utilized Lasso penalized logistic regression (Tibshirani, 1996) with leave-one-out cross-validation to select robust GAG properties that are most predictive of the clinical outcome (i.e., melanoma compared to healthy subject). The score was subsequently designed as a ratio, where the numerator is the sum of the properties associated with melanoma and the denominator is the sum of the properties associated with the healthy state. Each term was normalized using the regression coefficients.

[0788] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was platelet-poor plasma are referred to as GAG blood scores.

[0789] Accuracy Metrics

[0790] We evaluated the performance of the GAG score in the binary classification of a sample as either melanoma or healthy at varying threshold scores by deriving the receiver-operating-characteristic (ROC) curves. We measured the accuracy of the GAG score as the area under the curve (AUC) of its ROC curve (AUC is 1 for a perfect classifier and 0.5 for a random classifier). We selected as a potential cut-off value for the GAG score the score for which the accuracy was maximum, i.e. a sample whose GAG score is above this cut-off value has the maximum probability of being melanoma and below this cut-off value has the maximum probability of being healthy. The ROC curves were calculated using the pROC R-package, while the optimal cut-off using the OptimalCutpoints R-package.

[0791] Results

[0792] We analyzed a total of 16 blood samples from patients with M, of which 12 had uveal melanoma and 4 had skin (cutaneous) melanoma (M). We further analyzed 14 blood samples from healthy volunteers (H). Clinical data for this cohort is reported in Table L. We quantified the profile of GAGs in the 30 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table M). We compared the value of each property in the GAG profile in the M versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between M and H subjects (Table M).

[0793] We condensed changes in the GAG profile in M vs. H subjects GAG scores. We derived one M GAG score (Melanoma scoring formula #1), based on the following formula using blood measurements:

[00018] $M .Math. .Math. GAG .Math. .Math. score = \frac{7}{10} [4 .Math. s .Math. .Math. 6 .Math. s .Math. .Math. CS] + 5 .Math. (\frac{[6 .Math. s .Math. .Math. CS]}{[6 .Math. s .Math. .Math. CS] + [4 .Math. s .Math. .Math. CS]}) + \frac{1}{10} [Tot .Math. .Math. CS] .Math. (1 + \frac{1}{[Tot .Math. .Math. HA]}) - \frac{[0 .Math. s .Math. .Math. HS]}{[0 .Math. s .Math. .Math. HS] + [Ns .Math. .Math. HS]} - \frac{7}{1000} .Math. ([2 .Math. s .Math. .Math. HS] + [0 .Math. s .Math. .Math. HS])$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and [Tot CS] is the total concentration of CS (in g/mL), and [Tot HA] is the total concentration of HA (in g/mL). We subsequently calculated the score in each sample and observed that M samples have recurrently elevated scores with respect to H samples (FIG. 17). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 0.973. We identified an optimal cut-off equal to 0.92 for this score. Using this cut-off, M subjects could be distinguished from healthy individuals with 93.3% accuracy, 87.3% sensitivity, and 100% specificity.

[0794] We also investigated whether variations of the above GAG score would also be effective in distinguishing M from H subjects. To this end, we derived an alternative M GAG score (melanoma scoring formula #2) based only on CS measurements:

[00019] $M .Math. .Math. GAG .Math. .Math. score .Math. .Math. 2 = \frac{4 [6 .Math. s .Math. .Math. CS] + 5 [Tot .Math. .Math. CS]}{100} + 5 .Math. (\frac{[6 .Math. s .Math. .Math. CS]}{[4 .Math. s .Math. .Math. CS] + [6 .Math. s .Math. .Math. CS]})$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and [Tot CS] is the total concentration of CS (in g/mL). Using this alternative scoring system, M samples had recurrently higher scores than H samples (FIG. 18). The AUC was found to be 0.933. We identified an optimal cut-off equal to 1.19 for this score. Using this cut-off, M subjects could be distinguished from healthy individuals with 93.3% accuracy, 81.2% sensitivity, and 100% specificity.

[0795] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between melanoma as opposed to healthy subjects, and thus that GAG properties may be useful in melanoma screening (e.g. diagnosis). These results further demonstrate that different GAG scores can be designed to discriminate between melanoma and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

REFERENCES

[0796] Coppa G V, Gabrielli O, Buzzega D, Zampini L, Galeazzi T, Maccari F, Bertino E, Volpi N (2011). Composition and structure elucidation of human milk glycosaminoglycans. Glycobiology 21, 295-303.

[0797] Gatto, F., Volpi, N., Nilsson, H., Nookaew, I., Maruzzo, M., Roma, A., Johansson, M. E., Stierner, U., Lundstam, S., Basso, U., et al. (2016). Glycosaminoglycan Profiling in Patients' Plasma and Urine Predicts the Occurrence of Metastatic Clear Cell Renal Cell Carcinoma. Cell Rep 15, 1822-1836.

[0798] Tibshirani, R. (1996). Regression shrinkage and selection via the Lasso. Journal of the Royal Statistical Society Series B-Methodological 58, 267-288.

[0799] Volpi N and Maccari F. (2005.sup.1). Glycosaminoglycans composition of the largefreshwater mollusc bivalve Anodonta anodonta. Biomacromolecules 6, 3174-3180.

[0800] Volpi N and Maccari F (2005.sup.2). Microdetermination of chondroitin sulfate in normal human plasma by fluorophore-assisted carbohydrate electrophore-sis (FACE). Clin Chim Acta 356, 125-133.

[0801] Volpi, N., Galeotti, F., Yang, B., and Linhardt, R. J. (2014). Analysis of glycosaminoglycan-derived, precolumn, 2-aminoacridone-labeled disaccharides with LC-fluorescence and LC-MS detection. Nature protocols 9, 541-558.

[0802] Volpi, N., and Linhardt, R. J. (2010). High-performance liquid chromatography-mass spectrometry for mapping and sequencing glycosaminoglycan-derived oligosaccharides. Nature protocols 5, 993-1004.

TABLE-US-00010 TABLE L Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics Melanoma Healthy Number of samples 16 14 Age [years] 65 (53-71) 40 (35-52) Female 62.5% 78.6% Cancer subtype Uveal melanoma 12 Skin melanoma 4

[0803] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between M and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table M shows a summary of the results and shows the mean value for GAG properties as measured in M vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table M also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00011 TABLE M Mean in Healthy Mean in ROPE % in GAG property Individuals Melanoma threshold ROPE 0 s CS blood 77.699 75.170 3.81 70.45 2 s CS blood 0.025 0.026 0.00 17.49 6 s CS blood 1.201 3.291 0.12 0.14 4 s CS blood 20.775 20.531 1.04 41.82 2 s6 s CS blood 0.074 0.178 0.02 8.63 2 s4 s CS blood 0.035 0.047 0.00 8.18 4 s6 s CS blood 0.092 0.150 0.01 2.13 Tris CS blood 0.008 0.010 0.00 18.63 4 s/6 s CS blood 18.281 8.641 0.66 0.01 6 s/0 s CS blood 0.015 0.042 0.00 0.30 4 s/0 s CS blood 0.270 0.279 0.01 31.54 Charge CS blood 0.225 0.255 0.01 19.01 Total HA blood 0.142 0.151 0.02 56.68 Total CS blood 10.171 17.566 0.72 0.10 Tris HS blood 0.203 0.233 0.01 13.77 Ns6 s HS blood 0.365 0.277 0.02 11.03 Ns2 s HS blood 0.523 0.599 0.03 11.88 Ns HS blood 28.949 33.939 1.59 14.73 2 s6 s HS blood 0.180 0.185 0.03 31.57 6 s HS blood 33.410 40.470 1.87 12.32 2 s HS blood 4.490 2.674 0.18 2.95 0 s HS blood 30.456 20.738 1.27 2.83 Charge HS blood 0.721 0.811 0.04 11.39 Total HS blood 1.872 2.882 0.12 3.13

EXAMPLE 4

[0804] In this study we measured circulating GAG levels in subjects presenting with colorectal cancer (CRC).

[0805] Material and Methods

[0806] Plasma Sample Collection

[0807] Blood samples were obtained from 10 patients with colorectal cancers. All patients had evidence of disease and treatment-nave at the time of sampling.

[0808] Whole blood was collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0809] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0810] Glycosaminoglycan Profile Determination

[0811] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0812] Biomarker Score Design

[0813] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0814] Biomarker score design was as described in Example 3, but concerned colorectal cancer versus healthy groups (instead of melamona versus healthy).

[0815] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0816] Accuracy Metrics

[0817] We evaluated the performance of the CRC GAG score as in Example 3, but classifying a sample as colorectal cancer or healthy (instead of melanoma or healthy).

[0818] Results

[0819] We analyzed a total of 10 blood samples from patients with CRC. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table 0). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table N. We compared the value of each property in the GAG profile in the CRC versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between CRC and H subjects (Table 0).

[0820] We condensed changes in the GAG profile in CRC vs. H subjects into GAG scores. We derived one CRC GAG score, based on the following formula using blood measurements:

[00020] $CRC .Math. .Math. GAG .Math. .Math. score = \frac{5}{2} .Math. (\frac{[6 .Math. s .Math. .Math. CS]}{[4 .Math. s .Math. .Math. CS] + [6 .Math. s .Math. .Math. CS]}) + \frac{2}{15} .Math. (\frac{[2 .Math. s .Math. .Math. 6 .Math. s .Math. .Math. CS]}{[2 .Math. s .Math. .Math. 6 .Math. s .Math. .Math. CS] + [2 .Math. s .Math. .Math. CS]}) + \frac{1}{300} [6 .Math. s .Math. .Math. HS] + \frac{1}{30} .Math. ([2 .Math. s .Math. .Math. HS] + Tot .Math. .Math. HS)$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and [Tot HS] is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that CRC samples have recurrently elevated scores with respect to H samples (FIG. 19). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 0.998. We identified an optimal cut-off equal to 0.74 for this score. Using this cut-off, CRC subjects could be distinguished from healthy individuals with 98.5% accuracy, 100% sensitivity, and 98.2% specificity.

[0821] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between colorectal cancer as opposed to healthy subjects, and thus that GAG properties may be useful in colorectal cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between colorectal cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00012 TABLE N Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics Colorectal cancer Healthy Number of samples 10 58 Age [years] 74 (70-78) 56 (47-63) Female 50% 65.5% TNM pathologic stage T T2 1 T3 4 T4 2 Not Available 3 TNM pathologic stage N N0 3 N1 3 N2 3 Not Available 1 TNM pathologic stage M M0 7 M1 2 Not Available 1 TNM stage I-III 7 IV 3

[0822] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between CRC and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table 0 shows a summary of the results and shows the mean value for GAG properties as measured in CRC vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table 0 also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00013 TABLE O Mean in Mean in Healthy Colorectal ROPE % in GAG property Individuals cancer threshold ROPE 0 s CS blood 79.184 74.490 3.7483 40.44 2 s CS blood 0.022 0.072 0.0056 0.04 6 s CS blood 1.079 4.019 0.1137 1.35 4 s CS blood 19.577 21.535 1.1014 25.81 2 s6 s CS blood 0.060 0.268 0.0152 0.15 2 s4 s CS blood 0.020 0.068 0.0048 0.01 4 s6 s CS blood 0.060 0.093 0.0068 4.36 Tris CS blood 0.008 0.025 0.0041 1.59 4 s/6 s CS blood 36.456 9.332 1.3002 0.00 6 s/0 s CS blood 0.013 0.033 0.0018 2.35 4 s/0 s CS blood 0.237 0.283 0.0156 17.51 Charge CS blood 0.199 0.269 0.0123 5.22 Total HA blood 0.178 0.132 0.0081 6.13 Total CS blood 5.444 8.719 0.3199 2.35 Tris HS blood 0.314 0.187 0.0327 9.26 Ns6 s HS blood 0.965 0.324 0.0533 0.06 Ns2 s HS blood 0.751 0.284 0.0566 0.05 Ns HS blood 18.970 16.824 1.1047 17.89 2 s6 s HS blood 0.486 0.215 0.0448 2.03 6 s HS blood 7.147 28.052 1.0077 0.37 2 s HS blood 4.444 16.120 0.3736 0.09 0 s HS blood 58.306 33.676 2.3275 0.43 Charge HS blood 0.478 0.682 0.0289 1.79 Total HS blood 0.225 1.204 0.0509 0.10

EXAMPLE 5

[0823] In this study we measured circulating GAG levels in subjects presenting with gastrointestinal neuroendocrine tumors (GNET).

[0824] Material and Methods

[0825] Plasma Sample Collection

[0826] Blood samples were obtained from 10 patients with a gastrointestinal neuroendocrine tumor. All patients had evidence of disease and treatment-nave at the time of sampling.

[0827] Whole blood was collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0828] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0829] Glycosaminoglycan Profile Determination

[0830] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0831] Biomarker Score Design

[0832] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0833] Biomarker score design was as described in Example 3, but concerned gastro-intestinal neuroendocrine tumor versus healthy groups (instead of melanoma versus healthy).

[0834] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0835] Accuracy Metrics

[0836] We evaluated the performance of the GNET GAG score as in Example 3, but classifying a sample as GNET or healthy (instead of melanoma or healthy).

[0837] Results

[0838] We analyzed a total of 10 blood samples from patients with GNET. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table Q). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table P. We compared the value of each property in the GAG profile in the GNET versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between GNET and H subjects (Table Q).

[0839] We condensed changes in the GAG profile in GNET vs. H subjects into GAG scores. We derived one GNET GAG score, based on the following formula using blood measurements:

GNET GAG score=3/4[2s6s CS]3/20[Tris CS]+2 Charge CS+1/30[2s HS]1/300[0s HS]

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and Charge CS is the weighted charge of CS. We subsequently calculated the score in each sample and observed that GNET samples have recurrently elevated scores with respect to H samples (FIG. 20). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.90 for this score. Using this cut-off, GNET subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0840] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between gastrointestinal neuroendocrine tumors as opposed to healthy subjects, and thus that GAG properties may be useful in gastrointestinal neuroendocrine tumour screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between gastrointestinal neuroendocrine tumors and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00014 TABLE P Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics GNET Healthy Number of samples 10 58 Age [years] 58 (49-71) 56 (47-63) Female 70% 65.5% ENETS Grade G1 7 G2 3

[0841] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between GNET and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table Q shows a summary of the results and shows the mean value for GAG properties as measured in GNET vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table Q also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00015 TABLE Q Mean in Healthy Mean in ROPE % in GAG property Individuals GNET threshold ROPE 0 s CS blood 79.184 66.570 3.7483 0.580 2 s CS blood 0.022 0.043 0.0056 5.740 6 s CS blood 1.079 3.577 0.1137 0.060 4 s CS blood 19.577 28.872 1.1014 0.453 2 s6 s CS blood 0.060 0.541 0.0152 0.013 2 s4 s CS blood 0.020 0.087 0.0048 0.007 4 s6 s CS blood 0.060 0.202 0.0068 0.367 Tris CS blood 0.008 0.027 0.0041 0.907 4 s/6 s CS blood 36.456 9.511 1.3002 0.000 6 s/0 s CS blood 0.013 0.050 0.0018 0.060 4 s/0 s CS blood 0.237 0.433 0.0156 0.387 Charge CS blood 0.199 0.345 0.0123 0.020 Total HA blood 0.178 0.146 0.0081 11.547 Total CS blood 5.444 6.139 0.3199 19.593 Tris HS blood 0.314 0.299 0.0327 17.340 Ns6 s HS blood 0.965 0.412 0.0533 0.480 Ns2 s HS blood 0.751 0.549 0.0566 12.900 Ns HS blood 18.970 27.823 1.1047 3.247 2 s6 s HS blood 0.486 0.263 0.0448 5.327 6 s HS blood 7.147 33.651 1.0077 0.020 2 s HS blood 4.444 9.908 0.3736 0.940 0 s HS blood 58.306 26.747 2.3275 0.000 Charge HS blood 0.478 0.755 0.0289 0.053 Total HS blood 0.225 1.376 0.0509 0.033

EXAMPLE 6

[0842] In this study we measured circulating GAG levels in subjects presenting a blood cancer, more specifically with chronic lymphoid leukemia (CLL).

[0843] Material and Methods

[0844] Plasma Sample Collection

[0845] Blood samples were obtained from 10 patients with chronic lymphoid leukemia. All patients had evidence of disease and treatment-nave at the time of sampling.

[0846] Whole blood was collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0847] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0848] Glycosaminoglycan Profile Determination

[0849] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0850] Biomarker Score Design

[0851] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0852] Biomarker score design was as described in Example 3, but concerned CLL versus healthy groups (instead of melanoma versus healthy).

[0853] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0854] Accuracy Metrics

[0855] We evaluated the performance of the CLL GAG scores as in Example 3, but classifying a sample as CLL or healthy (instead of melanoma or healthy).

[0856] Results

[0857] We analyzed a total of 10 blood samples from patients with CLL. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table S). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table R. We compared the value of each property in the GAG profile in the CLL versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between CLL and H subjects (Table S).

[0858] We condensed changes in the GAG profile in CLL vs. H subjects into GAG scores. We derived one CLL GAG score (CLL GAG score #1), based on the following formula using blood measurements:

CLL GAG score=2/15[6s CS]+2/3[2s6s CS][4s6s CS]+1/2Charge CS1/150[6s HS]1/250[0s HS]+2/15Tot HS

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and Tot HS is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that CLL samples have recurrently elevated scores with respect to H samples (FIG. 21). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 0.974. We identified an optimal cut-off equal to 0.25 for this score. Using this cut-off, CLL subjects could be distinguished from healthy individuals with 95.6% accuracy, 100% sensitivity, and 94.8% specificity.

[0859] We also investigated whether variations of the above GAG score would also be effective in distinguishing CLL from H subjects. To this end, we derived an alternative CLL GAG score (CLL GAG score #2) using only CS properties:

[00021] $CLL .Math. .Math. GAG .Math. .Math. score = \frac{1}{12} [6 .Math. s .Math. .Math. CS] + \frac{4}{7} [2 .Math. s6s .Math. .Math. CS] - [4 .Math. s6s .Math. .Math. CS] - \frac{1}{800} .Math. (\frac{[4 .Math. s .Math. .Math. CS]}{[6 .Math. s .Math. .Math. CS]})$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG. Using this alternative scoring system, CLL samples had recurrently higher scores than H samples (FIG. 22). The AUC was found to be 0.974. We identified an optimal cut-off equal to 0.23 for this score. Using this cut-off, CLL subjects could be distinguished from healthy individuals with 95.6% accuracy, 70% sensitivity, and 100% specificity.

[0860] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between CLL as opposed to healthy subjects, and thus that GAG properties may be useful in CLL screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between chronic lymphoid leukemia and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00016 TABLE R Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics CLL Healthy Number of samples 10 58 Age [years] 70 (65-73) 56 (47-63) Female 40% 65.5% Binet stage A + B 7 C 3

[0861] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between CLL and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table S shows a summary of the results and shows the mean value for GAG properties as measured in CLL vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table S also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00017 TABLE S Mean in Healthy Mean in ROPE % in GAG property Individuals CLL threshold ROPE 0 s CS blood 79.184 69.171 3.7483 4.967 2 s CS blood 0.022 0.033 0.0056 23.153 6 s CS blood 1.079 4.101 0.1137 0.187 4 s CS blood 19.577 25.599 1.1014 3.207 2 s6 s CS blood 0.060 0.520 0.0152 0.060 2 s4 s CS blood 0.020 0.064 0.0048 0.013 4 s6 s CS blood 0.060 0.206 0.0068 0.553 Tris CS blood 0.008 0.010 0.0041 80.113 4 s/6 s CS blood 36.456 8.101 1.3002 0.000 6 s/0 s CS blood 0.013 0.057 0.0018 0.147 4 s/0 s CS blood 0.237 0.373 0.0156 1.960 Charge CS blood 0.199 0.319 0.0123 0.353 Total HA blood 0.178 0.139 0.0081 6.187 Total CS blood 5.444 7.733 0.3199 2.267 Tris HS blood 0.314 0.286 0.0327 24.720 Ns6 s HS blood 0.965 0.461 0.0533 1.193 Ns2 s HS blood 0.751 0.581 0.0566 15.807 Ns HS blood 18.970 31.397 1.1047 0.620 2 s6 s HS blood 0.486 0.191 0.0448 0.933 6 s HS blood 7.147 25.152 1.0077 0.253 2 s HS blood 4.444 8.734 0.3736 2.033 0 s HS blood 58.306 28.838 2.3275 0.033 Charge HS blood 0.478 0.730 0.0289 0.087 Total HS blood 0.225 1.307 0.0509 0.007

EXAMPLE 7

[0862] In this study we measured circulating GAG levels in subjects presenting with bladder cancer (BCa).

[0863] Material and Methods

[0864] Urine Sample Collection

[0865] Urine samples were obtained from 6 patients with bladder cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0866] Urine was collected in polypropylene tubes and frozen at 80 C. until they were shipped for analysis in dry ice.

[0867] Glycosaminoglycan Profile Determination

[0868] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3, but on urine samples instead of blood samples.

[0869] Biomarker Score Design

[0870] A control group was formed using historical GAG profiles obtained from urine samples of 25 healthy individuals. These samples belonged to two distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0871] Biomarker design was as described in Example 3, but concerned bladder cancer versus healthy groups (instead of melanoma versus healthy).

[0872] Accuracy Metrics

[0873] We evaluated the performance of the BCa (bladder cancer) GAG score as in Example 3, but classifying a sample as bladder cancer or healthy (instead of melanoma or healthy).

[0874] Results

[0875] We analyzed a total of 6 urine samples from patients with BCa. We quantified the profile of GAGs in the 6 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table T). We further included as controls a group of 25 historical GAG profiles measured in urine samples of healthy individuals (H). We compared the value of each property in the GAG profile in the BCa versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between BCa and H subjects (Table T).

[0876] We condensed changes in the GAG profile in BCa vs. H subjects into GAG scores. We derived one BCa GAG score, based on the following formula using urine measurements:

BCa GAG score=3/4Charge CS+1/3[Tris HS]1/50[Ns HS]+1/250[2s HS]+3/4Tot HS

where terms in brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charged of CS, and Tot HS is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that BCa samples have recurrently elevated scores with respect to H samples (FIG. 23). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1.0. We identified an optimal cut-off equal to 0.92 for this score. Using this cut-off, BCa subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0877] These results indicate that there are measurable differences in GAG levels and/or chemical composition between bladder cancer as opposed to healthy subjects, and thus that GAG properties may be useful in bladder cancer screening (e.g. diagnosis).

[0878] These results further demonstrate that GAG scores can be designed to discriminate between bladder cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

[0879] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between BCa and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table T shows a summary of the results and shows the mean value for GAG properties as measured in BCa vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table T also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00018 TABLE T Mean in Mean in Healthy Bladder ROPE % in GAG property Individuals cancer threshold ROPE 0 s CS urine 34.963 24.071 1.645 2.558 2 s CS urine 0.076 0.120 0.005 0.175 6 s CS urine 21.566 26.563 1.129 3.992 4 s CS urine 40.519 47.111 2.104 9.483 2 s6 s CS urine 0.604 0.658 0.031 18.442 2 s4 s CS urine 0.458 0.566 0.024 10.242 4 s6 s CS urine 1.184 1.087 0.058 19.533 Tris CS urine 0.039 0.015 0.003 2.017 4 s/6 s CS urine 2.006 1.806 0.099 23.075 6 s/0 s CS urine 0.750 1.199 0.042 1.708 4 s/0 s CS urine 1.307 2.160 0.079 2.417 Charge CS urine 0.653 0.782 0.034 2.992 Total HA urine 0.070 0.054 0.003 9.417 Total CS urine 0.927 0.690 0.047 8.283 Tris HS urine 0.421 0.908 0.026 0.083 Ns6 s HS urine 1.195 0.995 0.077 8.325 Ns2 s HS urine 1.395 2.416 0.083 2.917 Ns HS urine 12.366 8.003 0.588 4.692 2 s6 s HS urine 1.041 0.831 0.056 12.725 6 s HS urine 6.880 6.158 0.354 13.250 2 s HS urine 3.557 9.911 0.269 2.000 0 s HS urine 71.768 66.827 3.528 31.158 Charge HS urine 0.334 0.377 0.017 16.833 Total HS urine 0.149 0.496 0.011 0.108

EXAMPLE 8

[0880] In this study we measured circulating GAG levels in subjects presenting with breast cancer (BC).

[0881] Material and Methods

[0882] Plasma Sample Collection

[0883] Blood samples were obtained from 10 patients with breast cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0884] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0885] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0886] Glycosaminoglycan Profile Determination

[0887] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0888] Biomarker Score Design

[0889] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0890] Biomarker score design was as described in Example 3, but concerned breast cancer versus healthy groups (instead of melanoma versus healthy).

[0891] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0892] Accuracy Metrics

[0893] We evaluated the performance of the BC GAG score as in Example 3, but classifying a sample as breast cancer or healthy (instead of melanoma or healthy).

[0894] Results

[0895] We analyzed a total of 10 blood samples from patients with BC. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table V). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table U. We compared the value of each property in the GAG profile in the BC versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between BC and H subjects (Table V).

[0896] We condensed changes in the GAG profile in BC vs. H subjects into GAG scores. We derived one BC GAG score, based on the following formula using blood measurements:

[00022] $BC .Math. .Math. GAG .Math. .Math. score = 5 .Math. (\frac{[6 .Math. s .Math. .Math. CS]}{[0 .Math. .Math. s .Math. .Math. CS]}) + \frac{5}{3} .Math. .Math. Charge .Math. .Math. CS + \frac{1}{50} .Math. Tot .Math. .Math. CS$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS, and Tot CS is the total concentration of CS (in g/mL). We subsequently calculated the score in each sample and observed that BC samples have recurrently elevated scores with respect to H samples (FIG. 24). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.94 for this score. Using this cut-off, BC subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0897] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between breast cancer as opposed to healthy subjects, and thus that GAG properties may be useful in breast cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between breast cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00019 TABLE U Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics BC Healthy Number of samples 10 58 Age [years] 52 (47-63) 56 (47-63) Female 100% 65.5%

[0898] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between BC and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table V shows a summary of the results and shows the mean value for GAG properties as measured in BC vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table V also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00020 TABLE V Mean in Healthy Mean in ROPE % in GAG property Individuals BC threshold ROPE 0 s CS blood 79.184 62.447 3.387 0.000 2 s CS blood 0.022 0.041 0.004 8.073 6 s CS blood 1.079 3.339 0.181 0.000 4 s CS blood 19.577 33.352 1.397 0.007 2 s6 s CS blood 0.060 0.037 0.012 26.493 2 s4 s CS blood 0.020 0.046 0.006 6.193 4 s6 s CS blood 0.060 0.135 0.008 3.060 Tris CS blood 0.008 0.018 0.004 8.760 4 s/6 s CS blood 36.456 10.086 0.885 0.000 6 s/0 s CS blood 0.013 0.054 0.003 0.000 4 s/0 s CS blood 0.237 0.542 0.024 0.007 Charge CS blood 0.199 0.378 0.016 0.000 Total HA blood 0.178 0.131 0.010 7.280 Total CS blood 5.444 16.779 0.495 0.000 Tris HS blood 0.314 0.315 0.032 20.673 Ns6 s HS blood 0.965 0.552 0.058 7.633 Ns2 s HS blood 0.751 0.396 0.061 2.440 Ns HS blood 18.970 28.124 1.135 4.313 2 s6 s HS blood 0.486 0.290 0.045 9.793 6 s HS blood 7.147 28.369 1.067 0.213 2 s HS blood 4.444 4.040 0.329 19.280 0 s HS blood 58.306 38.908 2.274 2.133 Charge HS blood 0.478 0.645 0.030 5.100 Total HS blood 0.225 1.491 0.061 0.053

EXAMPLE 9

[0899] In this study we measured circulating GAG levels in subjects presenting with ovarian cancer (OV).

[0900] Material and Methods

[0901] Plasma Sample Collection

[0902] Blood samples were obtained from 9 patients with ovarian cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0903] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0904] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0905] Glycosaminoglycan Profile Determination

[0906] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0907] Biomarker Score Design

[0908] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0909] Biomarker score design was as described in Example 3, but concerned ovarian cancer versus healthy groups (instead of melanoma versus healthy).

[0910] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0911] Accuracy Metrics

[0912] We evaluated the performance of the OV GAG score as in Example 3, but classifying a sample as ovarian cancer or healthy (instead of melanoma or healthy).

[0913] Results

[0914] We analyzed a total of 9 blood samples from patients with OV. We quantified the profile of GAGs in the 9 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table X). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table W. We compared the value of each property in the GAG profile in the OV versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between OV and H subjects (Table X).

[0915] We condensed changes in the GAG profile in OV vs. H subjects into GAG scores. We derived one OV GAG score, based on the following formula using blood measurements:

OV GAG score=2 Charge CS+1/50(Tot CSTotHS)+1/2Tot HA

where Charge CS is the weighted charge of CS, Tot CS is the total concentration of CS (in g/mL), Tot HS is the total concentration of HS (in g/mL), Tot HA is the total concentration of HA (in g/mL). We subsequently calculated the score in each sample and observed that OV samples have recurrently elevated scores with respect to H samples (FIG. 25). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.99 for this score. Using this cut-off, OV subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0916] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between ovarian cancer as opposed to healthy subjects, and thus that GAG properties may be useful in ovarian cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between ovarian cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00021 TABLE W Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics OV Healthy Number of samples 9 58 Age [years] 60 (54-65) 56 (47-63) Female 100% 65.5% FIGO stage I-II 5 III-IV 4

[0917] A metric or statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between OV and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table X shows a summary of the results and shows the mean value for GAG properties as measured in OV vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table X also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00022 TABLE X Mean in Healthy Mean in ROPE % in GAG property Individuals OV threshold ROPE 0 s CS blood 79.184 55.951 3.387 0.007 2 s CS blood 0.022 0.025 0.004 33.667 6 s CS blood 1.079 5.149 0.181 0.013 4 s CS blood 19.577 38.740 1.397 0.000 2 s6 s CS blood 0.060 0.102 0.012 4.427 2 s4 s CS blood 0.020 0.089 0.006 0.493 4 s6 s CS blood 0.060 0.124 0.008 0.573 Tris CS blood 0.008 0.034 0.004 5.333 4 s/6 s CS blood 36.456 6.880 0.885 0.000 6 s/0 s CS blood 0.013 0.068 0.003 0.033 4 s/0 s CS blood 0.237 0.591 0.024 0.067 Charge CS blood 0.199 0.451 0.016 0.000 Total HA blood 0.178 0.299 0.010 0.240 Total CS blood 5.444 18.854 0.495 0.000 Tris HS blood 0.314 0.396 0.032 18.673 Ns6 s HS blood 0.965 0.655 0.058 10.493 Ns2 s HS blood 0.751 0.756 0.061 18.793 Ns HS blood 18.970 12.247 1.135 6.920 2 s6 s HS blood 0.486 0.445 0.045 15.600 6 s HS blood 7.147 7.856 1.067 25.573 2 s HS blood 4.444 5.371 0.329 12.693 0 s HS blood 58.306 65.639 2.274 14.513 Charge HS blood 0.478 0.425 0.030 24.547 Total HS blood 0.225 0.703 0.061 0.867

EXAMPLE 10

[0918] In this study we measured circulating GAG levels in subjects presenting with a uterine cancer, more specifically endometrial cancer (EC).

[0919] Material and Methods

[0920] Plasma Sample Collection

[0921] Blood samples were obtained from 9 patients with endometrial cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0922] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0923] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0924] Glycosaminoglycan Profile Determination

[0925] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0926] Biomarker Score Design

[0927] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0928] Biomarker score design was as described in Example 3, but concerned endometrial cancer versus healthy groups (instead of melanoma versus healthy).

[0929] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0930] Accuracy Metrics

[0931] We evaluated the performance of the EC GAG score as in Example 3, but classifying a sample as endometrial cancer or healthy (instead of melanoma or healthy).

[0932] Results

[0933] We analyzed a total of 9 blood samples from patients with EC. We quantified the profile of GAGs in the 9 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table Z). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table Y. We compared the value of each property in the GAG profile in the EC versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between EC and H subjects (Table Z).

[0934] We condensed changes in the GAG profile in EC vs. H subjects into GAG scores. We derived one EC GAG score, based on the following formula using blood measurements:

EC GAG score=8/5Charge CS+2/15[6s CS]1.500[2s4s CS]+1/65Tot CS1/20Tot HS

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS, Tot CS is the total concentration of CS (in g/mL), and Tot HS is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that EC samples have recurrently elevated scores with respect to H samples (FIG. 26). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.97 for this score. Using this cut-off, EC subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0935] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between endometrial cancer as opposed to healthy subjects, and thus that GAG properties may be useful in endometrial cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between endometrial cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00023 TABLE Y Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics EC Healthy Number of samples 9 58 Age [years] 61 (59.5-67.5) 56 (47-63) Female 100% 65.5% FIGO stage I-II 5 III-IV 4

[0936] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between EC and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table Z shows a summary of the results and shows the mean value for GAG properties as measured in EC vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table Z also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00024 TABLE Z Mean in Healthy Mean in ROPE % in GAG property Individuals EC threshold ROPE 0 s CS blood 79.184 57.236 3.387 0.000 2 s CS blood 0.022 0.043 0.004 7.780 6 s CS blood 1.079 5.060 0.181 0.000 4 s CS blood 19.577 37.511 1.397 0.000 2 s6 s CS blood 0.060 0.112 0.012 7.433 2 s4 s CS blood 0.020 0.070 0.006 0.267 4 s6 s CS blood 0.060 0.138 0.008 0.647 Tris CS blood 0.008 0.015 0.004 25.200 4 s/6 s CS blood 36.456 6.918 0.885 0.000 6 s/0 s CS blood 0.013 0.073 0.003 0.000 4 s/0 s CS blood 0.237 0.593 0.024 0.013 Charge CS blood 0.199 0.438 0.016 0.007 Total HA blood 0.178 0.278 0.010 3.327 Total CS blood 5.444 13.506 0.495 0.080 Tris HS blood 0.314 0.414 0.032 16.500 Ns6 s HS blood 0.965 0.626 0.058 10.907 Ns2 s HS blood 0.751 0.728 0.061 25.093 Ns HS blood 18.970 19.273 1.135 15.727 2 s6 s HS blood 0.486 0.562 0.045 18.073 6 s HS blood 7.147 11.596 1.067 14.800 2 s HS blood 4.444 3.648 0.329 18.593 0 s HS blood 58.306 59.019 2.274 20.867 Charge HS blood 0.478 0.445 0.030 23.753 Total HS blood 0.225 0.705 0.061 0.387

EXAMPLE 11

[0937] In this study we measured circulating GAG levels in subjects presenting with a uterine cancer, more specifically cervical cancer (CST).

[0938] Material and Methods

[0939] Plasma Sample Collection

[0940] Blood samples were obtained from 9 patients with cervical cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0941] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0942] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0943] Glycosaminoglycan Profile Determination

[0944] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0945] Biomarker Score Design

[0946] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0947] Biomarker score design was as described in Example 3, but concerned cervical cancer versus healthy groups (instead of melanoma versus healthy).

[0948] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0949] Accuracy Metrics

[0950] We evaluated the performance of the CST GAG score as in Example 3, but classifying a sample as cervical cancer or healthy (instead of melanoma or healthy).

[0951] Results

[0952] We analyzed a total of 9 blood samples from patients with CST. We quantified the profile of GAGs in the 9 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table AB). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table AA. We compared the value of each property in the GAG profile in the CST versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between CST and H subjects (Table AB).

[0953] We condensed changes in the GAG profile in CST vs. H subjects into GAG scores. We derived one CST GAG score, based on the following formula using blood measurements:

CST GAG score=Charge CS+1/6[6s CS]+1/50(Tot CS+[Ns2s HS])1/4Tot HS

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS, Tot CS is the total concentration of CS (in g/mL), and Tot HS is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that CST samples have recurrently elevated scores with respect to H samples (FIG. 27). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.71 for this score. Using this cut-off, CST subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0954] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between cervical cancer as opposed to healthy subjects, and thus that GAG properties may be useful in cervical cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between cervical cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00025 TABLE AA Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics CST Healthy Number of samples 9 58 Age [years] 41 (36-47) 56 (47-63) Female 100% 65.5% FIGO stage I 4 II-IV 5

[0955] A metric of statistical confidence that any individual GAG property (CS, HS or or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between CST and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table AB shows a summary of the results and shows the mean value for GAG properties as measured in CST vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table AB also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00026 TABLE AB Mean in Healthy Mean in ROPE % in GAG property Individuals CST threshold ROPE 0 s CS blood 79.184 59.292 3.387 0.000 2 s CS blood 0.022 0.026 0.004 38.500 6 s CS blood 1.079 4.779 0.181 0.000 4 s CS blood 19.577 36.224 1.397 0.007 2 s6 s CS blood 0.060 0.086 0.012 22.693 2 s4 s CS blood 0.020 0.058 0.006 3.093 4 s6 s CS blood 0.060 0.110 0.008 2.173 Tris CS blood 0.008 0.027 0.004 12.913 4 s/6 s CS blood 36.456 7.374 0.885 0.000 6 s/0 s CS blood 0.013 0.069 0.003 0.007 4 s/0 s CS blood 0.237 0.572 0.024 0.000 Charge CS blood 0.199 0.412 0.016 0.000 Total HA blood 0.178 0.264 0.010 4.607 Total CS blood 5.444 13.299 0.495 0.000 Tris HS blood 0.314 0.465 0.032 9.513 Ns6 s HS blood 0.965 0.716 0.058 15.347 Ns2 s HS blood 0.751 2.068 0.061 0.600 Ns HS blood 18.970 17.607 1.135 14.773 2 s6 s HS blood 0.486 0.361 0.045 18.607 6 s HS blood 7.147 16.160 1.067 4.920 2 s HS blood 4.444 3.602 0.329 15.827 0 s HS blood 58.306 56.422 2.274 18.553 Charge HS blood 0.478 0.481 0.030 25.247 Total HS blood 0.225 0.803 0.061 0.833

EXAMPLE 12

[0956] In this study we measured circulating GAG levels in subjects presenting with a blood cancer, more specifically Non-Hodgkin lymphoma (NHL).

[0957] Material and Methods

[0958] Plasma Sample Collection

[0959] Blood samples were obtained from 10 patients with Non-Hodgkin lymphoma, specifically diffuse large B-cell lymphoma. All patients had evidence of disease and treatment-nave at the time of sampling.

[0960] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0961] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0962] Glycosaminoglycan Profile Determination

[0963] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0964] Biomarker Score Design

[0965] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0966] Biomarker score design was as described in Example 3, but concerned NHL versus healthy groups (instead of melanoma versus healthy).

[0967] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0968] Accuracy Metrics

[0969] We evaluated the performance of the NHL GAG score as in Example 3, but classifying a sample as NHL or healthy (instead of melanoma or healthy).

[0970] Results

[0971] We analyzed a total of 10 blood samples from patients with NHL. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table AD). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table AC. We compared the value of each property in the GAG profile in the NHL versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between NHL and H subjects (Table AD).

[0972] We condensed changes in the GAG profile in NHL vs. H subjects into GAG scores. We derived one NHL GAG score, based on the following formula using blood measurements:

NHL GAG score=5/6Charge CS+1/5[6s CS]1/7[Tris CS]+1/3Charge HS1/20Tot HS

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS, Charge HS is the weighted charge of HS, and Tot HS is the total concentration of HS (in g/mL). We subsequently calculated the score in each sample and observed that NHL samples have recurrently elevated scores with respect to H samples (FIG. 28). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.97 for this score. Using this cut-off, NHL subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0973] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between NHL as opposed to healthy subjects, and thus that GAG properties may be useful in NHL screening (e.g. diagnosis). These results demonstrate that GAG scores can be designed to discriminate between Non-Hodgkin lymphoma and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00027 TABLE AC Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics NHL Healthy Number of samples 10 58 Age [years] 66 (61-70.5) 56 (47-63) Female 60% 65.5% Ann Arbor Stage I-II 3 III-IV 7

[0974] A metric of statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between NHL and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table AD shows a summary of the results and shows the mean value for GAG properties as measured in NHL vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table AD also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00028 TABLE AD Mean in Healthy Mean in ROPE % in GAG property Individuals NHL threshold ROPE 0 s CS blood 79.184 62.083 3.387 0.000 2 s CS blood 0.022 0.039 0.004 6.153 6 s CS blood 1.079 4.503 0.181 0.000 4 s CS blood 19.577 33.104 1.397 0.013 2 s6 s CS blood 0.060 0.149 0.012 1.313 2 s4 s CS blood 0.020 0.088 0.006 0.140 4 s6 s CS blood 0.060 0.150 0.008 0.620 Tris CS blood 0.008 0.076 0.004 0.000 4 s/6 s CS blood 36.456 6.834 0.885 0.000 6 s/0 s CS blood 0.013 0.068 0.003 0.000 4 s/0 s CS blood 0.237 0.502 0.024 0.007 Charge CS blood 0.199 0.388 0.016 0.000 Total HA blood 0.178 0.247 0.010 5.253 Total CS blood 5.444 8.753 0.495 1.273 Tris HS blood 0.314 0.745 0.032 0.600 Ns6 s HS blood 0.965 0.728 0.058 17.687 Ns2 s HS blood 0.751 1.324 0.061 1.420 Ns HS blood 18.970 31.578 1.135 0.440 2 s6 s HS blood 0.486 0.324 0.045 18.307 6 s HS blood 7.147 17.917 1.067 1.700 2 s HS blood 4.444 4.217 0.329 19.887 0 s HS blood 58.306 39.880 2.274 0.280 Charge HS blood 0.478 0.656 0.030 0.773 Total HS blood 0.225 1.251 0.061 0.007

EXAMPLE 13

[0975] In this study we measured circulating GAG levels in subjects presenting with brain cancer, more specifically diffuse glioma (DG).

[0976] Material and Methods

[0977] Plasma Sample Collection

[0978] Blood samples were obtained from 11 patients with diffuse glioma, specifically 7 patients with astrocytoma, 3 with glioblastoma multiforme and 1 with oligoastrocytoma. All patients had evidence of disease and treatment-nave at the time of sampling.

[0979] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0980] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[0981] Glycosaminoglycan Profile Determination

[0982] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[0983] Biomarker Score Design

[0984] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[0985] Biomarker score design was as described in Example 3, but concerned diffuse glioma versus healthy groups (instead of melanoma versus healthy).

[0986] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[0987] Accuracy Metrics

[0988] We evaluated the performance of the DG GAG score as in Example 3, but classifying a sample as DG or healthy (instead of melanoma or healthy).

[0989] Results

[0990] We analyzed a total of 11 blood samples from patients with DG. We quantified the profile of GAGs in the 11 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table AF). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table AE. We compared the value of each property in the GAG profile in the DG versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between DG and H subjects (Table AF).

[0991] We condensed changes in the GAG profile in DG vs. H subjects into GAG scores. We derived one DG GAG score, based on the following formula using blood measurements:

DG GAG score=3/5Charge CS+1/7[6s CS]+1/50Total CS+1/2Total HA+1/1000[0s HS]

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS, and Total CS is the total concentration of CS (in g/mL), and Total HA is the total concentration of HA (in g/mL). We subsequently calculated the score in each sample and observed that DG samples have recurrently elevated scores with respect to H samples (FIG. 29). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 1 (perfect classifier). We identified an optimal cut-off equal to 0.95 for this score. Using this cut-off, DG subjects could be distinguished from healthy individuals with 100% accuracy, 100% sensitivity, and 100% specificity.

[0992] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between DG as opposed to healthy subjects, and thus that GAG properties may be useful in DG screening (e.g. diagnosis). These results further demonstrate that GAG scores (or GAG profiles) can be designed to discriminate between diffuse glioma and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00029 TABLE AE Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics DG Healthy Number of samples 11 58 Age [years] 66 (50.5-70.5) 56 (47-63) Female 54.5% 65.5% Glioma subtype Astrocytoma 7 Glioblastoma multiforme 3 Oligoastrocytoma 1

[0993] A metric or statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between DG and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table AF shows a summary of the results and shows the mean value for GAG properties as measured in DG vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table AF also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00030 TABLE AF Mean in Healthy Mean in ROPE % in GAG property Individuals DG threshold ROPE 0 s CS blood 79.184 57.146 3.387 0.000 2 s CS blood 0.022 0.037 0.004 13.253 6 s CS blood 1.079 5.269 0.181 0.000 4 s CS blood 19.577 36.417 1.397 0.000 2 s6 s CS blood 0.060 0.129 0.012 5.447 2 s4 s CS blood 0.020 0.077 0.006 0.053 4 s6 s CS blood 0.060 0.159 0.008 0.553 Tris CS blood 0.008 0.083 0.004 0.007 4 s/6 s CS blood 36.456 7.056 0.885 0.000 6 s/0 s CS blood 0.013 0.096 0.003 0.000 4 s/0 s CS blood 0.237 0.603 0.024 0.000 Charge CS blood 0.199 0.443 0.016 0.000 Total HA blood 0.178 0.307 0.010 0.860 Total CS blood 5.444 10.785 0.495 0.000 Tris HS blood 0.314 0.414 0.032 15.860 Ns6 s HS blood 0.965 0.614 0.058 10.020 Ns2 s HS blood 0.751 0.562 0.061 14.620 Ns HS blood 18.970 31.537 1.135 1.393 2 s6 s HS blood 0.486 0.289 0.045 13.933 6 s HS blood 7.147 18.620 1.067 1.820 2 s HS blood 4.444 5.401 0.329 11.213 0 s HS blood 58.306 32.310 2.274 0.413 Charge HS blood 0.478 0.747 0.030 0.507 Total HS blood 0.225 1.420 0.061 0.000

EXAMPLE 14

[0994] In this study we measured circulating GAG levels in subjects presenting with lung cancer (LC).

[0995] Material and Methods

[0996] Plasma Sample Collection

[0997] Blood samples were obtained from 10 patients with lung cancer, specifically non-small cell lung cancer. All patients had evidence of disease and treatment-nave at the time of sampling.

[0998] Whole blood tubes were collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[0999] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[1000] Glycosaminoglycan Profile Determination

[1001] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[1002] Biomarker Score Design

[1003] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[1004] Biomarker design was as described in Example 3, but concerned lung cancer versus healthy groups (instead of melanoma versus healthy).

[1005] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[1006] Accuracy Metrics

[1007] We evaluated the performance of the LC GAG score as in Example 3, but classifying a sample as lung cancer or healthy (instead of melanoma or healthy).

[1008] Results

[1009] We analyzed a total of 10 blood samples from patients with LC. We quantified the profile of GAGs in the 10 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table AH). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table AG. We compared the value of each property in the GAG profile in the LC versus H group. We assessed the statistical significance of changes for each GAG property between the two groups by calculating the highest density interval (HDI) for the mean difference using Bayesian estimation. The results showed marked differences in several GAG properties between LC and H subjects (Table AH).

[1010] We condensed changes in the GAG profile in LC vs. H subjects into GAG scores. We derived one LC GAG score (LC GAG score #1), based on the following formula using blood measurements:

[00023] $LC .Math. .Math. GAG .Math. .Math. score = \frac{1}{6} .Math. Total .Math. .Math. CS - (\frac{1}{30} .Math. \frac{[4 .Math. .Math. s .Math. .Math. CS]}{[6 .Math. s .Math. .Math. CS]} + \frac{1}{25} [0 .Math. .Math. s .Math. .Math. HS])$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, and Total CS is the total concentration of CS (in g/mL). We subsequently calculated the score in each sample and observed that LC samples have recurrently elevated scores with respect to H samples (FIG. 30). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 0.986. We identified an optimal cut-off equal to 0.83 for this score. Using this cut-off, LC subjects could be distinguished from healthy individuals with 97% accuracy, 90% sensitivity, and 98% specificity.

[1011] We also investigated whether variations of the above GAG score would also be effective in distinguishing LC from H subjects. To this end, we derived an alternative LC GAG score (LC GAG score #2) using only CS properties:

LC GAG score=5/4Charge CS+1/25Total CS

where Total CS is the total concentration of CS (in g/mL), and Charge CS is the weighted charge of CS. Using this alternative scoring system, LC samples had recurrently higher scores than H samples (FIG. 31). The AUC was found to be 0.997. We identified an optimal cut-off equal to 0.88 for this score. Using this cut-off, LC patients could be distinguished from healthy individuals with 98.5% accuracy, 90% sensitivity, and 100% specificity.

[1012] These results indicate that there are measurable differences in circulating GAG levels and/or chemical composition between lung cancer as opposed to healthy subjects, and thus that GAG properties may be useful in lung cancer screening (e.g. diagnosis). These results further demonstrate that GAG scores can be designed to discriminate between lung cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00031 TABLE AG Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics LC Healthy Number of samples 10 58 Age [years] 62.5 (58-64) 56 (47-63) Female 50% 65.5% TNM stage I-III 4 IV 6

[1013] A metric or statistical confidence that any individual GAG property (CS, HS or HA levels, individual disaccharide compositions, and charge HS or charge CS) displays a significant and practically appreciable alteration between LC and healthy samples was calculated. An alteration was defined as significant and practically appreciable if the difference in means between case and control falls within a so-called Region Of Practical Equivalence in less than 5% of all possible statistical distributions that can fit the data relative to a measurement. Table AH shows a summary of the results and shows the mean value for GAG properties as measured in LC vs. H subjects blood and Bayesian estimation for the statistical significance of the mean difference (% in ROPE, Region Of Practical Equivalence, where a % in ROPE >5 indicates that random distributions of values for the two groups have a practically equivalent mean in more than 5% of the cases, and it is therefore not significant). Table AH also shows results where certain ratios of individual types of disaccharide composition have been calculated. 4s/6s CS is the ratio of 4s CS to 6s CS. 6s/0s CS is the ratio of 6s CS to 0s CS (unsulfated CS). 4s/0s CS is the ratio of 4s CS to 0s CS (unsulfated CS).

TABLE-US-00032 TABLE AH Mean in Healthy Mean in ROPE % in GAG property Individuals LC threshold ROPE 0 s CS blood 79.184 67.533 3.387 0.600 2 s CS blood 0.022 0.043 0.004 4.407 6 s CS blood 1.079 2.570 0.181 0.113 4 s CS blood 19.577 30.433 1.397 0.247 2 s6 s CS blood 0.060 0.065 0.012 46.873 2 s4 s CS blood 0.020 0.043 0.006 1.773 4 s6 s CS blood 0.060 0.093 0.008 8.800 Tris CS blood 0.008 0.020 0.004 0.300 4 s/6 s CS blood 36.456 10.924 0.885 0.000 6 s/0 s CS blood 0.013 0.038 0.003 0.007 4 s/0 s CS blood 0.237 0.418 0.024 0.253 Charge CS blood 0.199 0.332 0.016 0.140 Total HA blood 0.178 0.158 0.010 25.273 Total CS blood 5.444 17.306 0.495 0.000 Tris HS blood 0.314 0.221 0.032 13.747 Ns6 s HS blood 0.965 0.326 0.058 0.753 Ns2 s HS blood 0.751 0.386 0.061 2.087 Ns HS blood 18.970 26.430 1.135 7.213 2 s6 s HS blood 0.486 0.370 0.045 23.787 6 s HS blood 7.147 40.953 1.067 0.013 2 s HS blood 4.444 4.067 0.329 17.167 0 s HS blood 58.306 17.681 2.274 0.000 Charge HS blood 0.478 0.842 0.030 0.000 Total HS blood 0.225 2.304 0.061 0.000

EXAMPLE 15

[1014] In this study we measured circulating GAG levels in subjects presenting with different types of cancers and compared them with GAG levels in healthy subjects.

[1015] Material and Methods

[1016] Plasma Sample Collection

[1017] Blood samples were obtained from 105 patients with different types of cancer. All patients had evidence of disease and treatment-nave at the time of sampling. The following cancer types were examined: breast cancer (BC), colorectal cancer (CRC), cervical cancer (CST), diffuse glioma (DG), endometrial cancer (EC), gastrointestinal neuroendocrine tumors (GNET), lymphoid leukemia (LL), Non-Hodgkin's lymphoma (NHL), lung cancer (LC), ovarian cancer (OV), prostate cancer (PCa).

[1018] Whole blood was collected in EDTA tubes and centrifuged to separate the plasma. Plasma was aliquoted into 220 uL vials and vials were frozen at 80 C. until they were shipped for analysis in dry ice.

[1019] Plasma GAGs correspond to blood GAGs. For the purpose of this example, plasma is therefore also interchangeably referred to as blood.

[1020] Glycosaminoglycan Profile Determination

[1021] Sample preparation and the measurement of GAG properties (GAG profile determination) was done as per Example 3.

[1022] Biomarker Score Design

[1023] A control group was formed using historical GAG profiles obtained from blood samples of 58 healthy individuals. These samples belonged to four distinct historical cohorts. The GAG profile was measured as described in the previous section.

[1024] The difference for each GAG property between each cancer type vs. the control group was described in terms of mean and standard deviation. The statistical significance in the change of each GAG property between groups was reported in previous Examples.

[1025] To verify whether changes in the GAG profile in cancer patients could be condensed into GAG scores, that can be used to differentiate cancer patients from healthy subjects, and therefore would be suitable as a cancer diagnostic, we utilized Lasso penalized logistic regression (Tibshirani, 1996, supra) with leave-one-out cross-validation to select robust GAG properties that are most predictive of the clinical outcome (i.e., cancer compared to healthy subject). The score was subsequently designed as a ratio, where the numerator is the sum of the properties associated with cancer and the denominator is the sum of the properties associated with the healthy state. Each term was normalized using the regression coefficients.

[1026] For the purpose of this study and for reasons elsewhere stated, scores computed from blood GAGs regardless if the processed sample was plasma are referred to as GAG blood scores.

[1027] Accuracy Metrics

[1028] We evaluated the performance of the GAG score in the binary classification of a sample as either cancer or healthy at varying threshold scores by deriving the receiver-operating-characteristic (ROC) curves. We measured the accuracy of the GAG score as the area under the curve (AUC) of its ROC curve (AUC is 1 for a perfect classifier and 0.5 for a random classifier). We selected as a potential cut-off value for the GAG score the score for which the accuracy was maximum, i.e. a sample whose GAG score is above this cut-off value has the maximum probability of being cancer and below this cut-off value has the maximum probability of being healthy. The ROC curves were calculated using the pROC R-package, while the optimal cut-off using the OptimalCutpoints R-package.

[1029] Results

[1030] We analyzed a total of 105 blood samples from patients with cancer. We quantified the profile of GAGs in the 105 samples, comprising the level and disaccharide chemical composition of (i) chondroitin sulfate (CS), (ii) heparan sulfate (HS), and (iii) hyaluronic acid (HA), as previously described (Gatto et al., 2016 and above). In total, 24 different properties (including 19 independent properties) were measured as part of the GAG profile of the blood samples (listed in Table AJ). We further included as controls a group of 58 historical GAG profiles measured in blood samples of healthy individuals (H). Clinical data for this cohort is reported in Table Al. We reported the value of each property in the GAG profile in each cancer type versus H group in Table AJ. The results showed marked differences in several GAG properties between cancer and H subjects, as shown in FIG. 32-35. The statistical significance of each difference with the healthy group was assessed separately for each cancer type in previous Examples.

[1031] We condensed changes in the GAG profile in cancer vs. H subjects into GAG scores. We derived an alternative Cancer GAG score (Cancer GAG Score #2/Cancer Blood score #2) to the blood score designed in Example 2, based on the following formula using blood measurements:

[00024] $Cancer .Math. .Math. GAG .Math. .Math. score = \frac{10}{3} .Math. Charge .Math. .Math. CS + Total .Math. .Math. CS (\frac{6 .Math. s .Math. .Math. CS}{[4 .Math. s .Math. .Math. CS] + [6 .Math. s .Math. .Math. CS]}) + \log_{2} (1 + \frac{[2 .Math. s .Math. .Math. HS]}{[0 .Math. s .Math. .Math. HS]})$

where terms in square brackets represent the fraction (mass fraction) of the disaccharide for the corresponding GAG, Charge CS is the weighted charge of CS and Total CS is the total concentration of CS (in g/mL). We subsequently calculated the score in each sample and observed that cancer samples have recurrently elevated scores with respect to H samples (FIG. 36). The performance of the GAG score was evaluated using the receiver-operating-characteristic (ROC) curves, and the area under the curve (AUC) was found to be 0.986. We identified an optimal cut-off equal to 1.39 for this score. Using this cut-off, cancer subjects could be distinguished from healthy individuals with 95.1% accuracy, 97.1% sensitivity, and 91.4% specificity.

[1032] These results demonstrate that GAG scores (or GAG profiles) can be designed to discriminate between cancer and healthy subjects, and that their significant AUC permit fine-tuned optimal cut-off scores, in order to maximize sensitivity and/or specificity, depending on the intended diagnostic use.

TABLE-US-00033 TABLE AI Clinical data for the cohort analysed in this study. Results are presented as medians (25.sup.th, 75.sup.th percentile) or percent or counts. Missing values were omitted. Baseline characteristics Cancer Healthy Number of samples 105 58 Age [years] 64 (52-71) 56 (47-63) Female 66.7% 65.5% of which low- Cancer type Total stage/grade* Breast cancer (BC) 10 N.A. Colorectal cancer (CRC) 10 7 Cervical cancer (CST) 9 4 Diffuse glioma (DG) 11 8 Endometrial cancer (EC) 9 5 Gastrointestinal neuroendocrine 10 7 tumors (GNET) Lymphoid leukemia (LL) 10 7 Non-Hodgkin's lymphoma (NHL) 10 3 Lung cancer (LC) 10 4 Ovarian cancer (OV) 9 5 Prostate cancer (PCa) 7 5 *Low-stage/grade disease is defined as TNM stage I to III for CRC and LC, FIGO stage I for CST, FIGO stage I to II for EC and OV, non-Grade IV glioma for DG, ENETS grade G1 for GNET, Binet stage A or B for LL, Ann Arbor stage I to II for NHL, and Gleason grade < 7 for PCa (N.A. if not available).

TABLE-US-00034 TABLE AJ Mean value (+/ standard deviation) for each GAG property in the healthy group and in each cancer type group. GAG properties referring to composition are expressed as mass fractions (weight/weight %). H BC CRC CST DG EC CS Charge 0.20 +/ 0.05 0.37 +/ 0.07 0.29 +/ 0.14 0.43 +/ 0.12 0.46 +/ 0.10 0.46 +/ 0.13 CS Concentration 5.71 +/ 3.39 17.02 +/ 6.38 9.50 +/ 6.65 13.38 +/ 2.42 10.86 +/ 2.70 14.00 +/ 6.71 HA Concentration 0.18 +/ 0.10 0.13 +/ 0.04 0.14 +/ 0.07 0.27 +/ 0.15 0.34 +/ 0.20 0.34 +/ 0.26 HS Charge 0.49 +/ 0.24 0.64 +/ 0.27 0.67 +/ 0.22 0.48 +/ 0.27 0.74 +/ 0.29 0.45 +/ 0.25 HS Concentration 0.65 +/ 1.06 1.97 +/ 1.71 1.57 +/ 1.11 1.04 +/ 0.75 1.60 +/ 1.03 0.87 +/ 0.53 0 s CS 78.41 +/ 6.30 63.05 +/ 6.55 72.30 +/ 13.33 57.05 +/ 12.07 56.17 +/ 8.66 55.06 +/ 13.09 2 s CS 0.15 +/ 0.25 0.07 +/ 0.07 0.08 +/ 0.05 0.04 +/ 0.03 0.04 +/ 0.02 0.04 +/ 0.02 6 s CS 1.10 +/ 0.82 3.39 +/ 0.81 4.96 +/ 7.07 5.53 +/ 2.63 5.58 +/ 1.89 6.18 +/ 3.29 4 s CS 19.88 +/ 5.28 33.13 +/ 6.25 21.75 +/ 6.10 37.01 +/ 9.75 36.74 +/ 6.87 38.18 +/ 10.08 2 s6 s CS 0.10 +/ 0.12 0.07 +/ 0.09 0.69 +/ 1.11 0.11 +/ 0.08 0.29 +/ 0.34 0.15 +/ 0.11 2 s4 s CS 0.11 +/ 0.27 0.10 +/ 0.15 0.08 +/ 0.05 0.09 +/ 0.07 0.60 +/ 1.55 0.08 +/ 0.06 4 s6 s CS 0.12 +/ 0.18 0.14 +/ 0.10 0.10 +/ 0.04 0.12 +/ 0.05 0.33 +/ 0.45 0.17 +/ 0.11 Tris CS 0.12 +/ 0.23 0.05 +/ 0.08 0.03 +/ 0.03 0.05 +/ 0.04 0.24 +/ 0.28 0.13 +/ 0.21 4 s/6 s CS 37.90 +/ 34.12 10.15 +/ 2.72 9.34 +/ 5.56 7.38 +/ 2.05 7.13 +/ 2.43 6.95 +/ 1.98 6 s/0 s CS 0.01 +/ 0.01 0.05 +/ 0.02 0.10 +/ 0.19 0.12 +/ 0.10 0.10 +/ 0.04 0.14 +/ 0.13 4 s/0 s CS 0.26 +/ 0.10 0.54 +/ 0.14 0.34 +/ 0.21 0.74 +/ 0.51 0.69 +/ 0.28 0.80 +/ 0.54 Charge CS 0.20 +/ 0.05 0.37 +/ 0.07 0.29 +/ 0.14 0.43 +/ 0.12 0.46 +/ 0.10 0.46 +/ 0.13 Tris HS 0.87 +/ 1.30 0.47 +/ 0.41 0.26 +/ 0.28 0.54 +/ 0.37 0.43 +/ 0.24 0.54 +/ 0.49 Ns6 s HS 1.51 +/ 1.67 0.72 +/ 0.64 0.40 +/ 0.32 0.85 +/ 0.51 3.61 +/ 8.35 0.66 +/ 0.32 Ns2 s HS 1.54 +/ 1.85 0.72 +/ 0.77 0.55 +/ 0.59 2.04 +/ 1.32 1.95 +/ 3.63 0.82 +/ 0.41 Ns HS 19.27 +/ 12.38 28.25 +/ 15.96 16.87 +/ 10.61 17.86 +/ 16.66 31.50 +/ 16.47 19.54 +/ 16.00 2 s6 s HS 1.35 +/ 1.87 0.79 +/ 0.79 0.25 +/ 0.20 0.42 +/ 0.33 0.76 +/ 1.34 0.89 +/ 0.89 6 s HS 12.92 +/ 15.49 25.64 +/ 16.22 30.86 +/ 21.82 18.24 +/ 16.19 22.27 +/ 18.51 14.98 +/ 13.66 2 s HS 4.55 +/ 3.48 4.29 +/ 3.55 16.56 +/ 11.65 3.72 +/ 3.27 6.47 +/ 9.07 3.65 +/ 2.28 0 s HS 58.00 +/ 21.93 39.16 +/ 28.64 34.26 +/ 22.11 56.36 +/ 28.65 33.00 +/ 26.25 58.93 +/ 24.52 Charge HS 0.48 +/ 0.24 0.64 +/ 0.27 0.67 +/ 0.22 0.48 +/ 0.27 0.74 +/ 0.29 0.45 +/ 0.25 GNET LL NHL LC OV PCa CS Charge 0.35 +/ 0.09 0.32 +/ 0.11 0.40 +/ 0.10 0.36 +/ 0.16 0.47 +/ 0.15 0.28 +/ 0.04 CS Concentration 6.19 +/ 2.67 7.73 +/ 2.50 9.01 +/ 3.29 17.26 +/ 6.07 20.32 +/ 11.30 7.67 +/ 3.00 HA Concentration 0.15 +/ 0.07 0.14 +/ 0.05 0.37 +/ 0.45 0.16 +/ 0.06 0.29 +/ 0.10 0.16 +/ 0.08 HS Charge 0.75 +/ 0.16 0.73 +/ 0.14 0.66 +/ 0.12 0.84 +/ 0.08 0.42 +/ 0.17 0.73 +/ 0.20 HS Concentration 1.58 +/ 0.89 1.41 +/ 0.60 1.24 +/ 0.54 2.85 +/ 2.13 0.77 +/ 0.49 1.37 +/ 0.53 0 s CS 66.26 +/ 9.15 68.90 +/ 10.13 60.40 +/ 9.85 64.39 +/ 15.56 53.19 +/ 15.24 72.56 +/ 3.39 2 s CS 0.04 +/ 0.02 0.04 +/ 0.02 0.04 +/ 0.02 0.05 +/ 0.03 0.03 +/ 0.02 0.04 +/ 0.02 6 s CS 3.64 +/ 1.77 4.27 +/ 2.96 5.43 +/ 2.89 3.88 +/ 4.38 7.03 +/ 4.57 2.76 +/ 1.37 4 s CS 29.11 +/ 8.52 25.83 +/ 7.59 33.43 +/ 7.04 31.32 +/ 11.49 39.28 +/ 11.00 23.93 +/ 2.03 2 s6 s CS 0.57 +/ 0.31 0.67 +/ 0.66 0.22 +/ 0.20 0.11 +/ 0.11 0.14 +/ 0.11 0.45 +/ 0.25 2 s4 s CS 0.09 +/ 0.05 0.07 +/ 0.02 0.14 +/ 0.10 0.07 +/ 0.05 0.14 +/ 0.14 0.05 +/ 0.02 4 s6 s CS 0.23 +/ 0.16 0.21 +/ 0.15 0.24 +/ 0.21 0.13 +/ 0.10 0.15 +/ 0.08 0.19 +/ 0.10 Tris CS 0.06 +/ 0.09 0.01 +/ 0.01 0.09 +/ 0.08 0.06 +/ 0.09 0.04 +/ 0.04 0.03 +/ 0.02 4 s/6 s CS 9.70 +/ 5.42 8.11 +/ 3.71 6.82 +/ 1.78 10.87 +/ 3.82 6.86 +/ 2.35 10.33 +/ 4.47 6 s/0 s CS 0.06 +/ 0.03 0.07 +/ 0.05 0.10 +/ 0.09 0.10 +/ 0.20 0.17 +/ 0.16 0.04 +/ 0.02 4 s/0 s CS 0.46 +/ 0.20 0.40 +/ 0.18 0.59 +/ 0.27 0.62 +/ 0.65 0.88 +/ 0.55 0.33 +/ 0.04 Charge CS 0.35 +/ 0.09 0.32 +/ 0.11 0.40 +/ 0.10 0.36 +/ 0.16 0.47 +/ 0.15 0.28 +/ 0.04 Tris HS 0.43 +/ 0.36 0.32 +/ 0.21 0.79 +/ 0.51 0.31 +/ 0.27 0.87 +/ 0.81 0.38 +/ 0.22 Ns6 s HS 0.47 +/ 0.36 0.47 +/ 0.34 0.90 +/ 0.57 0.47 +/ 0.45 2.10 +/ 2.79 0.92 +/ 0.71 Ns2 s HS 0.59 +/ 0.36 0.61 +/ 0.39 1.69 +/ 1.18 0.46 +/ 0.32 2.23 +/ 2.74 0.69 +/ 0.38 Ns HS 27.84 +/ 12.35 31.15 +/ 11.45 31.57 +/ 8.71 26.48 +/ 22.16 12.56 +/ 12.77 22.97 +/ 15.65 2 s6 s HS 0.25 +/ 0.14 0.21 +/ 0.15 1.54 +/ 3.68 0.39 +/ 0.25 1.70 +/ 2.37 0.99 +/ 2.12 6 s HS 33.31 +/ 16.49 29.41 +/ 20.90 19.22 +/ 12.72 49.81 +/ 30.15 9.48 +/ 8.57 30.67 +/ 24.46 2 s HS 10.03 +/ 6.16 8.77 +/ 6.07 4.31 +/ 3.19 4.19 +/ 3.91 5.52 +/ 4.70 12.96 +/ 6.62 0 s HS 27.11 +/ 16.26 29.10 +/ 15.07 40.02 +/ 11.18 17.79 +/ 8.86 65.57 +/ 15.90 30.46 +/ 21.78 Charge HS 0.75 +/ 0.16 0.73 +/ 0.14 0.66 +/ 0.12 0.84 +/ 0.08 0.42 +/ 0.17 0.73 +/ 0.20

CANCER BIOMARKERS

Inventors

Cpc classification

Classification Explorer

A61B10/0041

HUMAN NECESSITIES

Classification Explorer

G01N33/68

PHYSICS

Classification Explorer

G01N33/57488

PHYSICS

Classification Explorer

G01N33/57407

PHYSICS

Classification Explorer

G01N2400/40

PHYSICS

Classification Explorer

G01N2800/52

PHYSICS

Classification Explorer

G01N33/66

PHYSICS

International classification

Classification Explorer

G01N33/574

PHYSICS

Classification Explorer

G01N33/68

PHYSICS

Abstract

Claims

Description