Analytical Methods and Arrays for Use in the Same

Abstract

The present invention relates to an in vitro method for identifying agents capable of inducing respiratory sensitization in a mammal and arrays and diagnostic kits for use in such methods. In particular, the methods include measurement of the expression of the biomarkers listed in Table 1A, Table 1B and/or Table 1C in MUTZ-3 cells exposed to a test agent.

Claims

1. A method for identifying agents capable of inducing respiratory sensitization in a mammal comprising or consisting of the steps of: a) exposing a population of dendritic cells or a population of dendritic-like cells to a test agent; and b) measuring in the cells the expression of one or more biomarker(s) selected from the group defined in Table 1A and/or Table 1B; wherein the expression in the cells of the one or more biomarkers measured in step (b) is indicative of the sensitizing effect of the sample to be tested.

2. The method according to claim 1 further comprising: c) exposing a separate population of the dendritic cells or dendritic-like cells to one or more negative control agent that is not a respiratory sensitizer in a mammal; and d) measuring in the cells the expression of the one or more biomarker(s) measured in step (b) wherein the test agent is identified as a respiratory sensitizer in the event that the presence and/or amount in the test sample of the one or more biomarker measured in step (d) differs from the presence and/or amount in the control sample of the one or more biomarker measured in step (b).

3. The method according to claim 1 or 2 further comprising: e) exposing a separate population of the dendritic cells or dendritic-like cells to one or more positive control agent that is a respiratory sensitizer in a mammal; and f) measuring in the cells the expression of the one or more biomarker(s) measured in step (b) wherein the test agent is identified as a respiratory sensitizer in the event that the presence and/or amount in the test sample of the one or more biomarker measured in step (f) corresponds to the presence and/or amount in the positive control sample of the one or more biomarker measured in step (b).

4. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of one or more biomarkers defined in Table 1A, for example, at least 2 of the biomarkers defined in Table 1A.

5. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of OR5B21.

6. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of SLC7A7.

7. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of OR5B21 and SLC7A7.

8. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of one or more of the biomarkers defined in Table 1B, for example, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 or 13 of the biomarkers defined in Table 1B.

9. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of all of the biomarkers defined in Table 1B.

10. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of one or more of the biomarkers defined in Table 10, for example, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 274, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286 or 287 of the biomarkers defined in Table 10.

11. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of all of the biomarkers defined in Table 1C.

12. The method according to any one of the preceding claims wherein step (b) comprises or consists of measuring the expression of all of the biomarkers defined in Table 1.

13. The method according to any one of the preceding claims wherein step (b) comprises measuring the expression of a nucleic acid molecule encoding the one or more biomarker(s).

14. The method according to claim 13 wherein the nucleic acid molecule is a cDNA molecule or an mRNA molecule.

15. The method according to claim 13 wherein the nucleic acid molecule is an mRNA molecule.

16. The method according to claim 13 wherein the nucleic acid molecule is an cDNA molecule.

17. The method according to any one of claims 13 to 16 wherein measuring the expression of the one or more biomarker(s) in step (b) is performed using a method selected from the group consisting of Southern hybridisation, Northern hybridisation, polymerase chain reaction (PCR), reverse transcriptase PCR (RT-PCR), quantitative real-time PCR (qRT-PCR), nanoarray, microarray, macroarray, autoradiography and in situ hybridisation.

18. The method according to any one of claims 13 to 17 wherein measuring the expression of the one or more biomarker(s) in step (b) is determined using a DNA microarray.

19. The method according to any one of the preceding claims wherein measuring the expression of the one or more biomarker(s) in step (b) is performed using one or more binding moieties, each capable of binding selectively to a nucleic acid molecule encoding one of the biomarkers identified in Table 1.

20. The method according to claim 19 wherein the one or more binding moieties each comprise or consist of a nucleic acid molecule.

21. The method according to claim 20 wherein the one or more binding moieties each comprise or consist of DNA, RNA, PNA, LNA, GNA, TNA or PMO.

22. The method according to claim 19 or 20 wherein the one or more binding moieties each comprise or consist of DNA.

23. The method according to any one of claims 20 to 23 wherein the one or more binding moieties are 5 to 100 nucleotides in length.

24. The method according to any one of claims 20 24 wherein the one or more nucleic acid molecules are 15 to 35 nucleotides in length.

25. The method according to any one of claims 20 to 24 wherein the binding moiety comprises a detectable moiety.

26. The method according to claim 25 wherein the detectable moiety is selected from the group consisting of: a fluorescent moiety; a luminescent moiety; a chemiluminescent moiety; a radioactive moiety (for example, a radioactive atom); or an enzymatic moiety.

27. The method according to claim 26 wherein the detectable moiety comprises or consists of a radioactive atom.

28. The method according to claim 27 wherein the radioactive atom is selected from the group consisting of technetium-99m, iodine-123, iodine-125, iodine-131, indium-111, fluorine-19, carbon-13, nitrogen-15, oxygen-17, phosphorus-32, sulphur-35, deuterium, tritium, rhenium-186, rhenium-188 and yttrium-90.

29. The method according to claim 26 wherein the detectable moiety of the binding moiety is a fluorescent moiety.

30. The method according to any one of claims 1 to 21 wherein step (b) comprises or consists of measuring the expression of the protein of the one or more biomarker defined in step (b).

31. The method according to claim 30 wherein measuring the expression of the one or more biomarker(s) in step (b) is performed using one or more binding moieties each capable of binding selectively to one of the biomarkers identified in Table 1.

32. The method according to claim 31 wherein the one or more binding moieties comprise or consist of an antibody or an antigen-binding fragment thereof.

33. The method according to claim 32 wherein the antibody or fragment thereof is a monoclonal antibody or fragment thereof.

34. The method according to claim 32 or 33 wherein the antibody or antigen-binding fragment is selected from the group consisting of intact antibodies, Fv fragments (e.g. single chain Fv and disulphide-bonded Fv), Fab-like fragments (e.g. Fab fragments, Fab′ fragments and F(ab).sub.2 fragments), single variable domains (e.g. V.sub.H and V.sub.L domains) and domain antibodies (dAbs, including single and dual formats [i.e. dAb-linker-dAb]).

35. The method according to claim 34 wherein the antibody or antigen-binding fragment is a single chain Fv (scFv).

36. The method according to claim 31 wherein the one or more binding moieties comprise or consist of an antibody-like binding agent, for example an affibody or aptamer.

37. The method according to any one of claims 31 to 36 wherein the one or more binding moieties comprise a detectable moiety.

38. The method according to claim 37 wherein the detectable moiety is selected from the group consisting of a fluorescent moiety, a luminescent moiety, a chemiluminescent moiety, a radioactive moiety and an enzymatic moiety.

39. The method according to any one of the preceding claims wherein step (b) is performed using an array.

40. The method according to claim 39 wherein the array is a bead-based array.

41. The method according to claim 40 wherein the array is a surface-based array.

42. The method according to any one of claims 39 to 41 wherein the array is selected from the group consisting of: macroarray; microarray; nanoarray.

43. An array for use in a method according any one of the preceding claims, the array comprising one or more first binding agents as defined in any one of claims 19 to 29 and 31 to 38.

44. An array according to claim 43 comprising binding agents which are collectively capable of binding to all of the biomarkers defined in Table 1.

45. An array according to claim 43 or 44 wherein the first binding agents are immobilised.

46. The method according to any one of the preceding claims for identifying agents capable of inducing a respiratory hypersensitivity response.

47. The method according to any one of the preceding claims wherein the hypersensitivity response is a humoral hypersensitivity response.

48. The method according to claim 46 or 47 wherein the hypersensitivity response is a type I hypersensitivity response.

49. The method according to any one of the preceding claims for identifying agents capable of inducing respiratory allergy.

50. The method according to any one of the preceding claims wherein the population of dendritic cells or population of dendritic-like cells is a population of dendritic-like cells.

51. The method according to claim 50 wherein the dendritic-like cells are myeloid dendritic-like cells.

52. The method according to claim 51 wherein the myeloid dendritic-like cells are derived from myeloid dendritic cells.

53. The method according to claim 52 wherein the cells derived from myeloid dendritic cells are myeloid leukaemia-derived cells.

54. The method according to claim 53 wherein the myeloid leukaemia-derived cells are selected from the group consisting of KG-1, THP-1, U-937, HL-60, Monomac-6, AML-193 and MUTZ-3.

55. The method according to any one of the preceding claims wherein the dendritic-like cells are MUTZ-3 cells.

56. The method according to any one of the preceding claims wherein the one or more negative control agent provided in step (c) is selected from the group consisting of 1-butanol, 4-aminobenzoic acid, chlorobenzene, dimethyl formamide, ethyl vanillin, isopropanol, methyl salicylate, propylene glycol, potassium permanganate, Tween 80™ (polyoxyethylene (20) sorbitan monooleate) and zinc sulphate.

57. The method according to claim 56 wherein at least 2 control non-sensitizing agents are provided, for example, at least 3, 4, 5, 6, 7, 8, 9, 10 or at least 11 control non-sensitizing agents.

58. The method according to any one of the preceding claims wherein the one or more positive control agent provided in step (e) comprises or consists of one or more agent selected from the group consisting of ammonium hexachloroplatinate, ammonium persulfate, glutaraldehyde, hexamethylen diisocyanate, maleic anhydride, methylene diphenol diisocyanate, phtalic anhydride, toluendiisocyanate and trimellitic anhydride.

59. The method according to claim 58 wherein at least 2 control sensitizing agents are provided, for example, at least 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 or at least 20 control sensitizing agents.

60. The method according to any one of the preceding claims wherein the method is indicative of the sensitizing potency of the sample to be tested.

61. An array for use in a method according any one of the preceding claims, the array comprising one or more binding moieties as defined in any one of claims 19 to 29 and 31-38.

62. An array according to claim 61 wherein the binding moieties are capable of binding to all of the biomarkers defined in Table 1A.

63. An array according to claim 61 or 62 wherein the binding moieties are capable of binding to all of the biomarkers defined in Table 1B.

64. An array according to claim 61, 62 or 63 wherein the binding moieties are capable of binding to all of the biomarkers defined in Table 1C.

65. An array according to any one of claims 61 to 64 wherein the binding moieties are capable of binding to all of the biomarkers defined in Table 1.

66. An array according to any on of claims 61 to 64 wherein the binding moieties are immobilised.

67. Use of two or more biomarkers selected from the group defined in Table 1 in combination for identifying respiratory hypersensitivity response sensitising agents.

68. The use according to claim 67 wherein all of the biomarkers defined in Table 1 are used collectively for identifying hypersensitivity response sensitising agents.

69. An analytical kit for use in a method according any one of claims 1 to 60 comprising: A) an array according to any one of claims 61 to 66; and B) instructions for performing the method as defined in any one of claims 1 to 60 (optional).

70. An analytical kit according to claim 76 further comprising one or more control samples.

71. An analytical kit according to claim 69 comprising one or more non-sensitizing agent(s).

72. An analytical kit according to claim 69, 70 or 71 comprising one or more sensitizing agent(s).

73. A method or use substantially as described herein.

74. An array or kit substantially as described herein.

Description

[0180] Preferred, non-limiting examples which embody certain aspects of the invention will now be described, with reference to the following figures:

[0181] FIG. 1: Backward elimination of potential biomarkers for respiratory sensitization. 1029 genes, selected by p-value sorting, were used as input. After elimination of 727 genes, a local minimum in KLD was observed. Thus, the remaining 302 genes collectively hold the most information relevant for separating respiratory sensitizers from non-sensitizers. This biomarker signature was termed GARD Respiratory Prediction Signature.

[0182] FIG. 2: Principal component analysis based on 302 transcripts chosen by p-value filtering and backward elimination. A complete separation between samples stimulated with respiratory sensitizers (blue) and non-sensitizers (green) is observed.

[0183] FIGS. 3A and 3B: Estimation of the predictive power of the GARD Respiratory Prediction Signature using cross-validation. 20 Validation Biomarker Signatures were constructed using 70% of randomly chosen data (train set). The Validation Biomarker Signatures were subsequently used to classify the samples in the remaining 30% of the data (test set). FIG. 3A: ROC AUC distribution following SVM predictions of samples in the test set. FIG. 3B: Representative representation of prediction performance illustrated with principal component analysis.

[0184] FIG. 4: CD86 expression of MUTZ-3 cells following chemical stimulations. Data shown is an average of (chemical stimulations, n=3, DMSO and unstimulated cells (n=6), with error bars showing standard deviation. Statistical significance was determined by student's t-test, comparing each stimulation with its corresponding vehicle, with p<0.05 indicated by *.

[0185] FIGS. 5A, 5B, and 5C: Establishment of a predictive biomarker signature. FIG. 5A: Principal component analysis based on 1029 transcripts chosen by p-value filtering. FIG. 5B: Principal component analysis based on 302 transcripts chosen by Backward Elimination. Samples are colored as respiratory sensitizers (blue, n=27) or non-sensitizers (green, n=47). All data consisting of 74 samples, including all replicates, is represented. FIG. 5C: Respiratory sensitizers are colored according their mechanistic subdomain.

[0186] FIGS. 6A, 6B, and 6C: Estimation of the predictive power of the GRPS using an external test set and cross-validation. FIG. 6A: An external data set consisting of triplicates of non-sensitizers were mapped into the PCA space constructed by the GRPS. Only the train data are allowed to influence the principal components. All replicates of samples are represented (train data n=74, test data n=48). FIG. 6B: A Validation Biomarker Signature was constructed using 70% of randomly chosen data (train set). The train set was used to build a PCA space using the Validation Biomarker Signature as a variable input. The remaining 30% of data (test set) was mapped into this space without being allowed to influence the principal components. FIG. 6C: The random division of data into train set and test set were iterated 20 times. The ROC AUC distribution is reported with a box plot, with jittered data points overlaid. The median ROC AUC was 0.84.

EXAMPLES

Introduction

[0187] Respiratory sensitization to low-molecular weight compounds is a common cause of occupational asthma, which has been associated with fatal outcomes. To prevent the occurrence of respiratory chemical sensitizers and minimize risks in working environments, efforts are being made to develop assays that will predict a compound's′ ability to induce respiratory sensitization. However, to date no validated in vitro or in vivo method, in vitro or in vivo, exists that reliably accomplishes accurate classifications of chemicals as respiratory sensitizers. Recently, we presented a novel in vitro assay for assessment of skin sensitizers, called GARD (Johansson et al., 2011, BMC Genomics, 12:339). We have expanded the applicability of GARD to be able to also classify respiratory sensitizers, using a new genomic biomarker signature set comprising 302 genes associated with immunological events leading to maturation of dendritic cells. Thus, we present an assay with the combined ability to predict both skin and respiratory sensitization ability in assayed compounds.

Materials and Methods

Chemicals

[0188] A panel of 20 chemical compounds, consisting of 9 respiratory sensitizers and 11 non-sensitizers were used for cell stimulations. The sensitizers were glutaraldehyde, ammonium persulfate, phtalic anhydride, methylene diphenol diisocyanate, ammonium hexachloroplatinate, trimellitic anhydride, hexamethylen diisocyanate, maleic anhydride and toluendiisocyanate. The non-sensitizers were chlorobenzene, zinc sulphate, 4-aminobenzoic acid, methyl salicylate, ethyl vanillin, isopropanol, dimethyl formamide, 1-butanol, potassium permanganate, propylene glycol and tween 80 (Table 2). All chemicals were from Sigma-Aldrich, St. Louis, Mo., USA. Compounds were dissolved in either dimethyl sulfoxide (DMSO) or distilled water. Prior to stimulations, the cytotoxicity of all compounds was monitored, using propidium iodide (PI) (BD Biosciences, San Diego, Calif.) using protocol provided by the manufacturer. The relative viability of stimulated cells was calculated as:

[00001]$\mathrm{Relative}.Math..Math.\mathrm{viability}=\frac{\mathrm{fraction}.Math..Math.\mathrm{of}.Math..Math.\mathrm{viable}.Math..Math.\mathrm{stimulated}.Math..Math.\mathrm{cells}}{\mathrm{fraction}.Math..Math.\mathrm{of}.Math..Math.\mathrm{viable}.Math..Math.\mathrm{unstimulated}.Math..Math.\mathrm{cells}}.Math.100$

[0189] For toxic compounds, the concentration yielding 90% relative viability (Rv90) was used. For non-toxic compounds, a concentration of 500 μM was used. For non-toxic compounds that were insoluble at 500 μM in medium, the highest soluble concentration was used. For compounds dissolved in DMSO, the final concentration of DMSO in each well was 0.1%. The concentrations used for any given chemical are termed the ‘GARD input concentration’, and are listed in Table 2.

Chemical Exposure of the Cells

[0190] The human myeloid leukemia-derived cell line MUTZ-3 (DSMZ, Braunschweig, Germany) was maintained in α-MEM (Thermo Scientific Hyclone, Logan, Utah) supplemented with 20% (volume/volume) fetal calf serum (Invitrogen, Carlsbad, Calif.) and 40 ng/ml rhGM-CSF (Bayer HealthCare Pharmaceuticals, Seattle, Wash.), as described (Johansson H, Lindstedt M, Albrekt A S, Borrebaeck C A: A genomic biomarker signature can predict skin sensitizers using a cell-based in vitro alternative to animal tests. BMC Genomics 2011, 12:399; Rasaiyaah J, Yong K, Katz D R, Kellam P, Chain B M: Dendritic cells and myeloid leukaemias: plasticity and commitment in cell differentiation. Br J Haematol 2007, 138(3):281-290). Cultures were maintained at 200.000 cells/ml during expansion, with a media change every 3-4 days. No differentiating steps were performed and instead, the proliferating progenitor MUTZ-3 was used for stimulations. Prior to each experiment, the cells were immunophenotyped using flow cytometry as a quality control. Cells were seeded in 6-well plates at 200.000 cells/ml. Stock solutions of each compound were prepared in either DMSO or distilled water, and were subsequently diluted so the in-well concentrations corresponded to the GARD input concentration, and in-well concentrations of DMSO were 0.1%. Cells were incubated for 24 h at 37° C. and 5% CO.sub.2. Thereafter, cells were harvested and analyzed by flow cytometry. In parallel, harvested cells were lysed in TRIzol reagent (Invitrogen) and stored at −20° C. until RNA extraction. Stimulations with chemicals were performed in three individual experiments, so that triplicates samples were obtained.

Phenotypic Analysis with Flow Cytometry

[0191] All cell surface staining and washing steps were performed in PBS containing 1% BSA (w/v). Cells were incubated with specific mouse mAbs for 15 min at 4° C. The following mAbs were used for flow cytometry: FITC-conjugated CD1a (DakoCytomation, Glostrup, Denmark), CD34, CD86, and HLA-DR (BD Biosciences), PE-conjugated CD14 (DakoCytomation), CD54 and CD80 (BD Biosciences). Mouse IgG1, conjugated to FITC or PE were used as isotype controls (BD Biosciences) and PI was used to assess cell viability. FACSDiva software was used for data acquisition with FACSCanto II instrument (BD Bioscience). 10,000 events were acquired and gates were set based on light scatter properties to exclude debris and nonviable cells. Further data analysis was performed using FCS Express V3 (De Novo Software, Los Angeles, Calif.).

Phenotypic Analysis, Chemical Exposure, Cell Harvest and RNA Isolation

[0192] The maintenance and chemical stimulation of MUTZ-3 and all subsequent isolation of RNA and preparation of cDNA was performed as previously described (Johansson H, Albrekt A S, Borrebaeck C A K, Lindstedt M (2012) The GARD assay for assessment of chemical skin sensitizers. Toxicol in Vitro). In short, a phenotypic control of MUTZ-3 was performed prior to chemical stimulation. Stimulated cells were harvested and RNA was isolated. A control of the maturity state of the cells was performed by flow cytometric analysis of CD86. Preparation of cDNA and hybridization, washing and scanning of the Human Gene 1.0 ST Arrays (Affymetrix, Santa Clara, Calif., USA) was performed, according to standardized protocols provided by the manufacturer (Affymetrix).

Microarray Data Analysis and Statistical Methods

[0193] The method by which a predictive signature was established has been previously described (Johansson H, Lindstedt M, Albrekt A S, Borrebaeck C A (2011) A genomic biomarker signature can predict skin sensitizers using a cell-based in vitro alternative to animal tests. BMC Genomics 12: 399). In short, microarray data were normalized and quality checked with the RMA algorithm, using Affymetrix Expression Console (Affymetrix). The top 1029 predictors were selected by p-values from an ANOVA, comparing respiratory sensitizers and non-sensitizers. An algorithm for Backward Elimination (Johansson et al., 2011, supra.; Carlsson A, Wingren C, Kristensson M, Rose C, Ferno M, et al. (2011) Molecular serum portraits in patients with primary breast cancer predict the development of distant metastases. Proc Natl Acad Sci U S A 108: 14252-14257) was applied on the top 1029 predictors, to further reduce the biomarker signature size. The Backward Elimination algorithm was modified to minimize the Kullback-Leibler error (Kullback S, Leibler R A (1951) On Information and Sufficiency. Annals of Mathematical Statistics 22: 79-86) rather than maximizing the Area Under the Receiver Operating Characteristic (ROC AUC) (Lasko T A, Bhagwat J G, Zou K H, Ohno-Machado L (2005) The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform 38: 404-415), in order to enable continued signature optimization in cases where the ROC AUC reaches 1.0. The selected top 302 predictors were collectively designated “GARD Respiratory Prediction Signature” (GRPS). The script for Backwards Eliminations was programmed in R (R Development Core Team (2008) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria), with the additional package e1071 (Weingart S N, Iezzoni L I, Davis R B, Palmer R H, Cahalane M, et al. (2000) Use of administrative data to find substandard care: validation of the complications screening program. Med Care 38: 796-806). ANOVA analyses and visualization of results with Principal Component Analysis (PCA) (Ringner M (2008) What is principal component analysis? Nat Biotechnol 26: 303-304) were performed using Qlucore Omics Explorer 2.3 (Qlucore AB, Lund, Sweden). The predictive performance of the GRPS was estimated using an external dataset consisting of negative chemical stimulations, as well as a method for cross-validation based on Support Vector Machines (SVM) (Noble W S (2006) What is a support vector machine? Nat Biotechnol 24: 1565-1567), as described (Johansson et al., 2011, supra.). The biological relevance of the GRPS was explored using Ingenuity Pathways Analysis (IPA) (Ingenuity Systems, Inc. Mountain View, USA), by performing a ‘Core Analysis’. The top 1029 genes were used as IPA input along with fold change values. Biological relevance was established by exploring the Canonical Pathways associated with input molecules. The array data has been uploaded to ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) with accession number E-MEXP-3773.

Interrogation of the Method for Identification of the Prediction Signature

[0194] The data set was divided into a training set and a test set, consisting of 70% and 30%, of the chemical compounds, respectively. The division was performed randomly, while maintaining the proportions of sensitizers and non-sensitizers in each subset at the same ratio as in the complete data set. A test biomarker signature was identified in the training set, using ANOVA filtering and backward elimination, as described above. This test signature was used to train a Support Vector Machine (SVM) (Noble W S: What is a support vector machine? Nat Biotechnol 2006, 24(12):1565-1567), using the training set, which was thereafter applied to predict the samples of the test set. The process was repeated 20 times and the distribution of the area under the Receiver Operating Characteristic (ROC AUC) (Lasko T A, Bhagwat J G, Zou K H, Ohno-Machado L: The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform 2005, 38(5):404-415) was used as a measurement of the performance of the model.

Results

Analysis of the Transcriptional Profiles in Chemically Stimulated MUTZ-3 Cells

[0195] Following 24 h stimulations with a panel of reference chemicals, mRNA from MUTZ-3 was collected for transcriptional profiling. The stimulations included 9 different chemical respiratory sensitizers and 11 different non-sensitizers, all sampled in biological triplicates except for 4-aminobenzoic acid, which was sampled in 6 replicates due to internal controls, and potassium permanganate, which was sampled in only 2 replicates due to a faulty array. In addition, DMSO and distilled water was sampled in 6 replicates each, as vehicle controls. Summarized, the dataset ready for analysis consisted of 74 arrays, each with measurements of 29141 transcripts.

[0196] The first step of analysis involved a p-value filtering of the genes according to their ability to separate respiratory sensitizers from non-sensitizers, as determined by an ANOVA comparing the two groups. Based on previous experience, approximately 1000 genes is an appropriate amount of potential predictors to use as an input in an algorithm for backward elimination (Johansson H, Lindstedt M, Albrekt A S, Borrebaeck C A: A genomic biomarker signature can predict skin sensitizers using a cell-based in vitro alternative to animal tests. BMC Genomics 2011, 12:399.). Using a p-value cutoff at 0.0067 (FDR 19%), 1029 genes were identified. The backward elimination algorithm was applied, removing the predictor that contributes the least information in an iterative manner. A local minimum in Kullbach-Liebler Divergence (KLD) was observed when 727 predictors was eliminated (FIG. 1). The remaining 302 genes are collectively termed the “GARD Respiratory Prediction Signature”, and their ability to differentiate between respiratory chemical sensitizers and non-sensitizers are illustrated in FIG. 2.

Interrogation of the Analysis Used to Identify the Prediction Signature

[0197] To validate the predictive power of our signature, we used a machine learning method called the Support Vector Machine (SVM) (Noble, 2006, supra.), which maps the data from a training set in space in order to maximize the separation of gene expression induced by sensitizing and non-sensitizing chemicals. As a training set, 70% of the data set was chosen randomly and the entire process of biomarker selection was repeated. Starting with 29,141 transcripts, the signature was reduced to a gene list of equal size as the GARD Respiratory Prediction Signature, i.e. 302 transcripts, termed “Validation Biomarker Signature”, using ANOVA filtering and backward elimination, as described above. An SVM was trained on the train data, using the Validation Biomarker Signature. The trained SVM was then used to classify each sample in the remaining 30% of the data, i.e. the test set, as either a respiratory sensitizer or a non-sensitizer. The performance of the classifications was evaluated with the area under the Receiver Operating Characteristic (ROC AUC). This entire cross-validation was iterated 20 times, each time generating different train and test sets, with each train set yielding different Validation Biomarker Signatures. The results of these cross-validations are illustrated in FIG. 3. The median ROC AUC was found to be 0.84, with a range from 0.66 to 0.96. The large variations in predictive performance imply that the random exclusion of 30% of the data greatly affects the composition of the Validation Biomarker Signature. However, the ability to achieve ROC AUCs of up to 0.96 is strong evidence that when the model is trained on all available data, accurate classifications are indeed possible. This cross-validation demonstrates that the GARD Respiratory Prediction biomarkers are capable of accurately predicting respiratory sensitizing properties of unknown samples.

MUTZ-3 Phenotype in Unstimulated and Stimulated Cells

[0198] Prior to chemical challenge, the cells were quality controlled by measuring the cellular expression of common myeloid and dendritic cell markers using flow cytometry. These markers included CD1a, CD14, CD34, CD54, CD80, CD86 and HLA-DR. No deviations from previously published data were found (Johansson et al., 2011, supra.), ensuring that unstimulated cells were successfully maintained in an immature state. Following chemical stimulation, the general maturity state of the cells was controlled again, as determined by the expression of the co-stimulatory marker CD86, with results presented in FIG. 4. Upregulation of CD86 was evident after a number of chemical stimulations, however, due to large standard deviations, only glutaraldehyde and hexamethylen diisocyanate resulted in statistically significant upregulation of CD86. Furthermore, while not statistically significant, an upregulations of CD86 was also evident after a number of control stimulations. Thus, we concluded that CD86 was an unsuited biomarker for respiratory chemical sensitizers. However, many of the compounds used for stimulations in this study were poorly soluble in cell media, and could not be used in concentrations sufficient to induce cytotoxicity. To this end, the increase of CD86 expression can act as a complementary tool to ensure bioavailability of the chemical stimulations.

Analysis of the Transcriptional Profiles in Chemically Stimulated MUTZ-3 Cells

[0199] Following 24 h stimulations, with a panel of reference chemicals, mRNA from MUTZ-3 was collected for transcriptional profiling. The stimulations included 9 different chemical respiratory sensitizers and 11 different non-sensitizers (negative controls), all analyzed in biological triplicates except for 4-aminobenzoic acid, who was analyzed in 6 replicates due to internal controls, and potassium permanganate, which was analyzed in only 2 replicates due to a faulty array. In addition, DMSO and distilled water was analyzed in 6 replicates each, as vehicle controls. Summarized, the data set ready for analysis consisted of 74 arrays, each with measurements of 29,141 transcripts.

[0200] The first step of analysis involved a p-value filtering of the genes, according to their ability to discriminate respiratory sensitizers from non-sensitizers, as determined by an ANOVA comparing the two groups. Due to computational limits, approximately 1000 genes is an appropriate amount of potential predictors to use as an input in the algorithm for Backward Elimination. In the present data set, this pre-selection of predictor candidates resulted in 1029 genes, with a p-value of 0.0067 or lower, with a False Discovery Rate (FDR) (Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B 57: 289-300) of 19%. Collectively, these genes were able to separate respiratory sensitizers from non-sensitizers. However, a clear separation was not achieved, as illustrated with 3D Principal Component Analysis (PCA) (FIG. 5A). Reducing the number of predictors further, by the ranking given by their p-value, did not achieve a clear separation, even though the data contained predictor candidates with p-values down to 10.sup.−10. The Backward Elimination algorithm was then applied, removing the predictors (genes) that contribute the least information. A local minimum in Kullbach-Liebler Divergence (KLD) was observed when 727 predictors were eliminated (data not shown). The remaining 302 genes are collectively termed the “GARD Respiratory Prediction Signature” (GRPS), and their ability to differentiate between respiratory chemical sensitizers and non-sensitizers are illustrated in FIG. 5B. The identities of the genes are listed in Table 1.

[0201] Of note, there is a significantly larger variation of transcriptional profiles within the group of respiratory sensitizers, compared to the group of non-sensitizers. A similar phenomenon was observed also when studying skin sensitizers, which was related to the potency of the sensitizer, as well as the propensity of different chemicals to induce different signaling pathways (Johansson et al., 2011, supra.). However, categorically defined sensitizing potency is not available for these respiratory chemical sensitizers (Basketter D A, Kimber I (2011) Assessing the potency of respiratory allergens: uncertainties and challenges. Regul Toxicol Pharmacol 61: 365-372). Instead, we aimed to describe the differences in transcriptional profiles in relation to the mechanistic subdomain of each chemical sensitizer (Enoch S J, Roberts D W, Cronin M T (2010) Mechanistic Category Formation for the Prediction of Respiratory Sensitization. Chem Res Toxicol). FIG. 5C shows the same PCA plot as in FIG. 5B, with sensitizers colored according to their mechanistic subdomain, as listed in Table 2. Ammonium salts tend to be positioned further away from the cluster of non-sensitizers, indicating the most dissimilarity to non-sensitizers in terms of transcriptional profile. However, diisocyanates and acid anhydrides cluster closely together, leaving no possibility to draw any conclusion of any dissimilarities between these two groups at this point. To the best of our knowledge, glutaraldehyde has not been assigned to a mechanistic subdomain, although it groups closely with both acid anhydrides and diisocyanates, and these samples are thus denoted “Subdomain unknown” (FIG. 5C).

Evaluation of the Predictive Accuracy of the Prediction Signature

[0202] The predictive performance of the GRPS was evaluated in two ways. Firstly, an external test set consisting of non-sensitizers was used to confirm their position in a PCA plot, based on the GRPS. Secondly, we used a cross-validation method that randomly divided the data into training and test sets, which then were used to train and evaluate the Support Vector Machine classifications.

[0203] The first method was possible to perform due to the availability of an additional set of control chemicals, run in a previous set of experiments in which GARD was first conceived (Johansson et al., 2011, supra.). The compounds in this test set were benzaldehyde, chlorobenzene, diethyl phtalate, glycerol, lactic acid, octanoic acid, phenol, salicylic acid and sodium dodecyl sulphate, all sampled in biological triplicates. In addition, the test set contained nine samples of DMSO and unstimulated controls respectively. FIG. 6A shows the same PCA plot as FIG. 5B, in which the test set has been mapped based on the transcriptional profile of the samples, while not being allowed to influence the principal components. All samples of the test set are correctly grouped together with non-sensitizers of the train set. The lack of respiratory sensitizers in this test set was due to our reluctance to set any of these samples aside, when performing the analysis used to establish the GRPS. Any samples included in this analysis are inappropriate to include in a test set due to the risk of over fitting.

[0204] To overcome the problem of having no respiratory sensitizers in a true test set, we used a method for cross-validation. As a training set, 70% of the data set was chosen randomly and the entire process of biomarker selection was repeated. Starting with 29,141 transcripts, the signature was reduced to a gene list of equal size to the GRPS, i.e. 302 transcripts, termed “Validation Biomarker Signature”, using p-value filtering and Backward Elimination, as described above. A Support Vector Machine (SVM) (Noble WS (2006) What is a support vector machine? Nat Biotechnol 24: 1565-1567) was trained on the training data set, using the Validation Biomarker Signature. The trained SVM was then used to classify each sample in the remaining 30% of the data, i.e. the test data set, as either a respiratory sensitizer or a non-sensitizer. The performance of the classifications was evaluated with the area under the Receiver Operating Characteristic (ROC AUC). This entire cross-validation was iterated 20 times, each time generating different training and test sets, with each training set yielding different Validation Biomarker Signatures. The results of these cross-validations are illustrated in FIG. 6B-C. The median ROC AUC was found to be 0.84, with a range from 0.66 to 0.96. In addition, the Validation Call Frequency (VCF) for each gene in the GRPS is listed in Table 1. The VCF describes the frequency by which a certain gene has been included in any of the 20 Validation Biomarker Signatures, thus providing a second measurement by which the predictors can be ranked.

Canonical Pathways Associated with the GARD Respiratory Prediction Signature

[0205] Aiming to investigate the biologic response initiated by respiratory chemical sensitizers in MUTZ-3 cells, the data was analyzed with Ingenuity Pathway Analysis (IPA). The top 1029 genes, selected with p-value filtering, were used as input into IPA, along with values of fold change for each gene. Of the 1029 genes, IPA was able to map 933 to unique IDs. Taking duplicates into account, the dataset ready for IPA analysis consisted of 901 molecules. The primary objective was to elucidate which canonical pathways identified molecules are associated with. Results are listed in Table 1, in order of statistical significance according to IPA.

[0206] A clear majority of these identified and significantly regulated pathways are mainly driven by a limited set of molecules. These pathways include TREM1 signaling, altered T cell and B cell signaling in rheumatoid arthritis, communication between adaptive and innate immune cells, B cell development, aryl hydrocarbon receptor signaling, dendritic cell maturation, CD28 signaling in T-helper cells, lipid antigen presentation by CD1, cytotoxic T cell mediated apoptosis of target cells and autoimmune thyroid disease signaling. Of note, central for all of these pathways is the bridge between innate and adaptive immunity, and the engagement of innate immune responses initiated by recognition of foreign substances, leading to dendritic cell maturation. Key aspects of this process that is well monitored by the GRPS include upregulation of innate receptors, such as TLRs and AHR, upregulation of antigen presentation-associated molecules, such as HLA and CD1, upregulation of co-stimulatory molecules, such as CD86 and CD40, and upregulation of proinflammatory effector molecules, such as IL-8 and IL-1B.

Discussion

[0207] A variety of chemicals induce allergic sensitization of not only the skin, but also the respiratory tract, giving rise to occupational asthma and other symptoms (Kimber I, Dearman R J (1997) Cell and molecular biology of chemical allergy. Clin Rev Allergy Immunol 15: 145-168). While not as prevalent as chemicals inducing skin sensitization leading to allergic contact dermatitis, identification and hazard assessment of respiratory chemical sensitizers are equally important, not least due to the severe symptoms, with possible fatal outcomes (Chester D A, Hanna E A, Pickelman B G, Rosenman K D (2005) Asthma death after spraying polyurethane truck bedliner. Am J Ind Med 48: 78-84; Kimber I, Wilks M F (1995) Chemical respiratory allergy. Toxicological and occupational health issues. Hum Exp Toxicol 14: 735-736).

[0208] Recently, we presented a cell-based in vitro test method for skin sensitizers, called GARD, which is able to classify chemicals with high accuracy (Johansson et al., 2011, supra.; Johansson H, Albrekt A S, Borrebaeck C A K, Lindstedt M (2012) The GARD assay for assessment of chemical skin sensitizers. Toxicol in Vitro). The assay relies on the transcriptional profiling of MUTZ-3 cells following compound stimulation, using a predefined biomarker signature as readout. As measurements of these biomarkers are based on expression array technology, great opportunities exist to broaden the applicability domain of this assay. In the current study, we present a further development of GARD, allowing for the identification of respiratory chemical sensitizers, using a separate biomarker signature termed GARD Respiratory Prediction Signature (GRPS). The GRPS was identified, using a set of reference chemicals known to be either respiratory sensitizers or non-sensitizers, and identifying differentially expressed genes in these two groups by an ANOVA p-value filtering followed by a feature selection algorithm for Backward Elimination. The intended use of the obtained GRPS will thus be in a combined in vitro assay, in which MUTZ-3 cells are stimulated with unknown compounds to be classified. Using the two distinct biomarker signatures, the compound can be classified as a skin sensitizer, respiratory sensitizer or a non-sensitizer. Chemicals that are able to induce both respiratory and skin sensitization will also be specifically classified as such.

[0209] The predictive performance of the assay in classifying respiratory chemical allergens was estimated by two forms of validations. Firstly, an external test set consisting of triplicates of 9 negative stimulations were successfully classified, as shown in FIG. 6A. Secondly, a thorough approach of cross-validation was applied, in which 30% of the data was repeatedly excluded at random to form a test set that were later on classified with an SVM model trained on the remaining 70% of the data. Results of this cross-validation are presented as ROC AUCs, (FIGS. 6B and 6C) with a median of 0.84 in a range from 0.66 to 0.96. The large variations in predictive performance imply that the random exclusion of 30% of the data greatly affects the composition of the Validation Biomarker Signature. Indeed, the variation between different Validation Biomarker Signatures are larger and VCF:s are smaller than expected from previous experience (Johansson et al., 2011, supra.). The impact of the composition of each Validation Biomarker Signature has been investigated, and correlations were considered for among a number of factors such as the presence of certain mechanistic domains in the train set and number of replicates of each stimulation that were removed from each training set. No obvious patterns were revealed that could explain the variations in predictive performance. The deviations are thus believed to depend on the relatively low number of samples used for stimulation, which is in turn a consequence of the scarce number of reference compounds available. However, the ability to achieve ROC AUCs of up to 0.96 display strong evidence that when the model is trained on all available data, accurate classifications are indeed possible.

[0210] The current absence of validated or even widely accepted methods for hazard assessment of chemicals inducing respiratory sensitization is in large part due to the lack of understanding of the immunobiological mechanisms by which chemical respiratory sensitization occur (Isola D, Kimber I, Sarlo K, Lalko J, Sipes I G (2008) Chemical respiratory allergy and occupational asthma: what are the key areas of uncertainty? J Appl Toxicol 28: 249-253). Specifically, one of the most elusive issues yet to be resolved is the role of the IgE antibody in allergic sensitization of the respiratory tract to chemicals, and whether there are mechanisms through which such sensitization can be achieved that are independent of IgE antibody (Kimber I, Dearman R J (2002) Chemical respiratory allergy: role of IgE antibody and relevance of route of exposure. Toxicology 181-182: 311-315). There are indeed correlations between IgE antibody levels and clinical symptoms for a number of chemical allergens, e.g. for acid anhydrides. On the contrary, less than half of the patients that are sensitized to diisocyanates demonstrate specific IgE antibody in serum. Still, the consensus opinion is that the relationship between IgE antibody and chemical respiratory allergy is strong (Kimber I, Basketter D A, Gerberick G F, Ryan C A, Dearman R J (2011) Chemical allergy: translating biology into hazard characterization. Toxicol Sci 120 Suppl 1: S238-268). The most convincing argument is that there are technical difficulties in designing probes that successfully detect IgE antibodies specific for chemical haptens. In addition, the time of sampling of blood for allergen-specific IgE in relation to the last time of exposure might influence the outcome of such assays.

[0211] In an attempt to monitor and compare the transcriptional profiles of different subtypes of respiratory chemical allergens, FIG. 1C shows a PCA based on the GRPS genes, with chemicals colored according to mechanistic domain. No apparent difference is detectable between diisocyanates and acid anhydrides in this plot, as these two groups cluster closely together. While this does not resolve the issue of possibly different mechanistic pathways in sensitization in vivo, IgE dependent or IgE independent, it does confirm that these groups of chemicals induce similar transcriptional changes in MUTZ-3. Instead, the most extreme transcriptional changes are induced by ammonium salts, such as ammonium hexachloroplatinate and ammonium persulfate. However, the major differences in transcriptional profiles of these two compounds are detectable along the axis of the first principal component, i.e. in the same vectorial direction as sensitizers are separated from non-sensitizers. Thus, we conclude that the GRPS is capable of accurately classifying allergens from various mechanistic subdomains.

[0212] To further explore the biological effects of sensitizing chemicals on MUTZ-3, an IPA analysis was performed. In order to achieve sufficient significance in the data, the top 1029 genes from p-value filtering were used as input in the IPA software, rather than the top 302 genes of the GRPS. The IPA output presented in Table 1 lists the canonical signaling pathways with which the top 1029 genes are most significantly associated. A majority of these pathways are mainly driven by a core set of molecules, including CD86, CD40, TLR1, TLR6, various HLA-DR molecules and CD1 molecules. Thus, respiratory chemical sensitizers induce increased antigen presentation and upregulation of co-stimulatory molecules in MUTZ-3, arguably in response to ligation of various pattern recognition receptors (PRRs) and intracellular oxidative stress, as indicated by the significance of aryl hydrocarbon receptor (AHR) signaling and glutathione metabolism.

[0213] Taken together, the biologic response in MUTZ-3 to chemical respiratory allergens is dominated by innate immune response signaling pathways that ultimately leads to cell maturation of this dendritic cell model, with enhanced antigen presentation and interaction with other immune cells as the end result. Furthermore, novel findings of usage of signaling pathways that has previously been associated with respiratory sensitization to protein allergens shed some light on the biological process leading to sensitization of the respiratory tract in response to chemical allergens. Thus, the GRPS is indeed relevant in an immunologically mechanistic perspective, and provides measurement of transcripts that monitor the biologic events leading to respiratory sensitization.

[0214] In conclusion, we present a predictive biomarker signature for respiratory chemical sensitizers in MUTZ-3 cells that complement the previously described GARD assay for assessment of skin sensitizers. The ability to test for two different endpoints in the same sample provides an attractive and hitherto unique assay for safety assessment of chemicals in an in vitro environment.

References

[0215] 1. Boverhof D R, Billington R, Gollapudi B B, Hotchkiss J A, Krieger S M, et al. (2008) Respiratory sensitization and allergy: current research approaches and needs. Toxicol Appl Pharmacol 226: 1-13.

[0216] 2. Banks D E, Tarlo S M (2000) Important issues in occupational asthma. Curr Opin Pulm Med 6: 37-42.

[0217] 3. Sastre J, Vandenplas O, Park H S (2003) Pathogenesis of occupational asthma. Eur Respir J 22: 364-373.

[0218] 4. Zammit-Tabona M, Sherkin M, Kijek K, Chan H, Chan-Yeung M (1983) Asthma caused by diphenylmethane diisocyanate in foundry workers. Clinical, bronchial provocation, and immunologic studies. Am Rev Respir Dis 128: 226-230.

[0219] 5. Bernstein D I, Patterson R, Zeiss C R (1982) Clinical and immunologic evaluation of trimellitic anhydride-and phthalic anhydride-exposed workers using a questionnaire with comparative analysis of enzyme-linked immunosorbent and radioimmunoassay studies. J Allergy Clin Immunol 69: 311-318.

[0220] 6. Murdoch R D, Pepys J, Hughes E G (1986) IgE antibody responses to platinum group metals: a large scale refinery survey. Br J Ind Med 43: 37-43.

[0221] 7. Docker A, Wattie J M, Topping M D, Luczynska C M, Newman Taylor A J, et al. (1987) Clinical and immunological investigations of respiratory disease in workers using reactive dyes. Br J Ind Med 44:534-541.

[0222] 8. Bourne M S, Flindt M L, Walker J M (1979) Asthma due to industrial use of chloramine. Br Med J 2: 10-12.

[0223] 9. Verstraelen S, Bloemen K, Nelissen I, Witters H, Schoeters G, et al. (2008) Cell types involved in allergic asthma and their use in in vitro models to assess respiratory sensitization. Toxicol In Vitro 22: 1419-1431.

[0224] 10. Dearman R J, Basketter D A, Kimber I (1992) Variable effects of chemical allergens on serum IgE concentration in mice. Preliminary evaluation of a novel approach to the identification of respiratory sensitizers. J Appl Toxicol 12: 317-323.

[0225] 11. Dearman R J, Skinner R A, Humphreys N E, Kimber I (2003) Methods for the identification of chemical respiratory allergens in rodents: comparisons of cytokine profiling with induced changes in serum IgE. J Appl Toxicol 23: 199-207.

[0226] 12. Verstraelen S, Nelissen I, Hooyberghs J, Witters H, Schoeters G, et al. (2009) Gene profiles of THP-1 macrophages after in vitro exposure to respiratory (non-) sensitizing chemicals: identification of discriminating genetic markers and pathway analysis. Toxicol In Vitro 23: 1151-1162.

[0227] 13. Verstraelen S, Nelissen I, Hooyberghs J, Witters H, Schoeters G, et al. (2009) Gene profiles of a human bronchial epithelial cell line after in vitro exposure to respiratory (non-)sensitizing chemicals: identification of discriminating genetic markers and pathway analysis. Toxicology 255: 151-159.

[0228] 14. Verstraelen S, Nelissen I, Hooyberghs J, Witters H, Schoeters G, et al. (2009) Gene profiles of a human alveolar epithelial cell line after in vitro exposure to respiratory (non-)sensitizing chemicals: identification of discriminating genetic markers and pathway analysis. Toxicol Lett 185: 16-22.

[0229] 15. Lalko J F, Kimber I, Dearman R J, Gerberick G F, Sarlo K, et al. (2011) Chemical reactivity measurements: potential for characterization of respiratory chemical allergens. Toxicol In Vitro 25: 433-445.

[0230] 16. Johansson H, Lindstedt M, Albrekt A S, Borrebaeck C A (2011) A genomic biomarker signature can predict skin sensitizers using a cell-based in vitro alternative to animal tests. BMC Genomics 12: 399.

[0231] 17. Johansson H, Albrekt A S, Borrebaeck C A K, Lindstedt M (2012) The GARD assay for assessment of chemical skin sensitizers. Toxicol in Vitro.

[0232] 18. Santegoets S J, Masterson A J, van der Sluis P C, Lougheed S M, Fluitsma D M, et al. (2006) A CD34(+) human cell line model of myeloid dendritic cell differentiation: evidence for a CD14(+)CD11b(+) Langerhans cell precursor. J Leukoc Biol 80: 1337-1344.

[0233] 19. Masterson A J, Sombroek C C, De Gruijl T D, Graus Y M, van der Vliet H J, et al. (2002) MUTZ-3, a human cell line model for the cytokine-induced differentiation of dendritic cells from CD34+ precursors. Blood 100: 701-703.

[0234] 20. Larsson K, Lindstedt M, Borrebaeck C A (2006) Functional and transcriptional profiling of MUTZ-3, a myeloid cell line acting as a model for dendritic cells. Immunology 117: 156-166.

[0235] 21. Carlsson A, Wingren C, Kristensson M, Rose C, Ferno M, et al. (2011) Molecular serum portraits in patients with primary breast cancer predict the development of distant metastases. Proc Natl Acad Sci U S A 108: 14252-14257.

[0236] 22. Kullback S, Leibler R A (1951) On Information and Sufficiency. Annals of Mathematical Statistics 22: 79-86.

[0237] 23. Lasko T A, Bhagwat J G, Zou K H, Ohno-Machado L (2005) The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform 38: 404-415.

[0238] 24. R Development Core Team (2008) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria.

[0239] 25. Weingart S N, Iezzoni L I, Davis R B, Palmer R H, Cahalane M, et al. (2000) Use of administrative data to find substandard care: validation of the complications screening program. Med Care 38: 796-806.

[0240] 26. Ringner M (2008) What is principal component analysis? Nat Biotechnol 26: 303-304.

[0241] 27. Noble W S (2006) What is a support vector machine? Nat Biotechnol 24: 1565-1567.

[0242] 28. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B 57: 289-300.

[0243] 29. Basketter D A, Kimber I (2011) Assessing the potency of respiratory allergens: uncertainties and challenges. Regul Toxicol Pharmacol 61: 365-372.

[0244] 30. Enoch S J, Roberts D W, Cronin M T (2010) Mechanistic Category Formation for the Prediction of Respiratory Sensitization. Chem Res Toxicol.

[0245] 31. Kimber I, Dearman R J (1997) Cell and molecular biology of chemical allergy. Clin Rev Allergy Immunol 15: 145-168.

[0246] 32. Chester D A, Hanna E A, Pickelman B G, Rosenman K D (2005) Asthma death after spraying polyurethane truck bedliner. Am J Ind Med 48: 78-84.

[0247] 33. Kimber I, Wilks M F (1995) Chemical respiratory allergy. Toxicological and occupational health issues. Hum Exp Toxicol 14: 735-736.

[0248] 34. Isola D, Kimber I, Sarlo K, Lalko J, Sipes I G (2008) Chemical respiratory allergy and occupational asthma: what are the key areas of uncertainty? J Appl Toxicol 28: 249-253.

[0249] 35. Kimber I, Dearman R J (2002) Chemical respiratory allergy: role of IgE antibody and relevance of route of exposure. Toxicology 181-182: 311-315.

[0250] 36. Kimber I, Basketter D A, Gerberick G F, Ryan C A, Dearman R J (2011) Chemical allergy: translating biology into hazard characterization. Toxicol Sci 120 Suppl 1: S238-268.

TABLE-US-00001 TABLE 1 “Core”, “preferred” and “optional” biomarkers from the GARD Respiratory Prediction Signature. Affymetrix Validation Call Gene Symbol Entrez Gene ID Probe Set ID Frequency (A) Core biomarkers 1. OR5B21 ENST00000360374 7948330 100 2. SLC7A7 ENST00000404278 7977786 95 (B) Preferred biomarkers 3. PIP3-E ENST00000265198 8130408 85 4. BTNL8 ENST00000400706 8116537 85 5. CLEC4A ENST00000360500 7953723 90 6. HIST4H4 ENST00000358064 7961483 80 7. YKT6 ENST00000223369 8132580 80 8. FLJ32679 /// ENST00000327271 7981895 85 GOLGA8G /// GOLGA8E 9. PACSIN3 ENST00000298838 7947801 90 10. PDE1B ENST00000243052 7955943 80 11. NQO1 ENST00000320623 8002303 80 12. CAMK1D ENST00000378845 7926223 95 13. MYB ENST00000341911 8122202 95 14. — ENST00000387396 8065752 80 15. GRK5 ENST00000369106 7930894 90 (C) Optional biomarkers 16. CD86 ENST00000330540 8082035 100 17. CD1A ENST00000289429 7906339 85 18. WWOX ENST00000355860 7997352 85 19. IKZF2 ENST00000374319 8058670 85 20. FUCA1 ENST00000374479 7913694 80 21. C10orf76 ENST00000370033 7935951 80 22. AMICA1 ENST00000356289 7952022 80 23. PDPK2 /// PDPK1 ENST00000382326 7998825 80 24. AZU1 ENST00000334630 8024038 80 25. ACN9 ENST00000360382 8134415 80 26. PDPN ENST00000400804 7898057 75 27. LOC642587 NM_001104548 7909422 75 28. SEC61A2 ENST00000379051 7926189 75 29. ELA2 ENST00000263621 8024056 75 30. BMP2K ENST00000335016 8096004 75 31. HCCS ENST00000321143 8165995 75 32. CXorf26 ENST00000373358 8168447 75 33. TYSND1 ENST00000287078 7934114 70 34. CARS ENST00000380525 7945803 70 35. NECAP1 ENST00000339754 7953715 70 36. CDH26 ENST00000348616 8063761 70 37. SERPINB1 ENST00000380739 8123598 70 38. STEAP4 ENST00000301959 8140840 70 39. TXNIP ENST00000369317 7904726 65 40. — ENST00000386628 7925821 65 41. C12orf35 ENST00000312561 7954711 65 42. HMGA2 ENST00000393578 7956867 65 43. KRT16 ENST00000301653 8015376 65 44. GGTLC2 ENST00000215938 8071662 65 45. — ENST00000386437 8089926 65 46. OSBPL11 ENST00000393455 8090277 65 47. FAM71F1 ENST00000315184 8135945 65 48. ATP6V1B2 ENST00000276390 8144931 65 49. LOC128102 AF252254 7904429 60 50. TBX19 ENST00000367821 7907146 60 51. NID1 ENST00000264187 7925320 60 52. LPXN ENST00000263845 7948332 60 53. C15orf45 AK057017 7982375 60 54. RNF111 ENST00000380504 7983953 60 55. — ENST00000386861 7993183 60 56. CD33 ENST00000262262 8030804 60 57. TANK ENST00000259075 8045933 60 58. ANKRD44 ENST00000282272 8057990 60 59. WDFY1 ENST00000233055 8059361 60 60. SDC4 ENST00000372733 8066513 60 61. TMPRSS11B ENST00000332644 8100701 60 62. AFF4 ENST00000265343 8114083 60 63. HBEGF ENST00000230990 8114572 60 64. XK ENST00000378616 8166723 60 65. SLAMF7 ENST00000368043 7906613 55 66. S100A4 ENST00000368715 7920271 55 67. MPZL3 ENST00000278949 7952036 55 68. — GENSCAN00000044853 7967586 55 69. TRAV8-3 ENST00000390435 7973298 55 70. LOC100131497 GENSCAN00000046821 7980481 55 71. KIAA1468 ENST00000299783 8021496 55 72. SPHK2 ENST00000245222 8030078 55 73. — ENST00000309260 8096554 55 74. CCR6 ENST00000283506 8123364 55 75. GSTA3 ENST00000370968 8127087 55 76. RALA ENST00000005257 8132406 55 77. C7orf53 ENST00000312849 8135532 55 78. — AF480566 8141421 55 79. CERCAM ENST00000372842 8158250 55 80. — hsa-mir-147 8163729 55 81. NFYC ENST00000372655 7900468 50 82. CD53 ENST00000271324 7903893 50 83. PSEN2 ENST00000366783 7910146 50 84. CISD1 ENST00000333926 7927649 50 85. SCD ENST00000370355 7929816 50 86. MED19 ENST00000337672 7948293 50 87. SYT17 ENST00000396244 7993624 50 88. KRT16 /// ENST00000399124 8013465 50 LOC400578 /// MGC102966 89. C18orf51 ENST00000400291 8023864 50 90. CD79A ENST00000221972 8029136 50 91. C19orf56 ENST00000222190 8034448 50 92. AGFG1 ENST00000409979 8048847 50 93. FOXP1 ENST00000318796 8088776 50 94. TLR6 ENST00000381950 8099841 50 95. SUSD3 ENST00000375472 8156393 50 96. — ENST00000387842 8176921 50 97. — ENST00000387842 8177424 50 98. GPA33 ENST00000367868 7922029 45 99. CDC123 ENST00000281141 7926207 45 100. C10orf11 ENST00000354343 7928534 45 101. — ENST00000322493 7937971 45 102. PTMAP7 AF170294 7976239 45 103. ARRDC4 ENST00000268042 7986350 45 104. — ENST00000388199 7997738 45 105. — ENST00000388437 8009299 45 106. KRT9 ENST00000246662 8015357 45 107. — ENST00000379371 8035868 45 108. HDAC4 ENST00000345617 8060030 45 109. CD200 ENST00000315711 8081657 45 110. PAPSS1 ENST00000265174 8102214 45 111. ORAI2 ENST00000356387 8135172 45 112. — AK124536 8144569 45 113. ZBTB10 ENST00000379091 8147040 45 114. — ENST00000387422 8159963 45 115. RAB9A ENST00000243325 8166098 45 116. — — 7895613 40 117. DRD5 ENST00000304374 7905025 40 118. CNR2 ENST00000374472 7913705 40 119. OIT3 ENST00000334011 7928330 40 120. — ENST00000386981 7933008 40 121. C10orf90 ENST00000356858 7936996 40 122. OR52D1 ENST00000322641 7938008 40 123. ZNF214 ENST00000278314 7946288 40 124. — ENST00000386959 7954690 40 125. ART4 ENST00000228936 7961507 40 126. RCBTB2 ENST00000344532 7971573 40 127. HOMER2 ENST00000304231 7991034 40 128. WWP2 ENST00000359154 7996976 40 129. WDR24 ENST00000248142 7998280 40 130. MED31 ENST00000225728 8011968 40 131. CALM2 ENST00000272298 8052010 40 132. DLX2 ENST00000234198 8056784 40 133. BTBD3 ENST00000399006 8060988 40 134. — ENST00000339367 8075817 40 135. TBCA ENST00000380377 8112767 40 136. GIN1 ENST00000399004 8113403 40 137. NOL7 ENST00000259969 8116969 40 138. — ENST00000402365 8117628 40 139. C7orf28B /// ENST00000325974 8138128 40 C7orf28A 140. DPP7 ENST00000371579 8165438 40 141. hCG_1749005 NR_003933 8167640 40 142. PNPLA4 ENST00000381042 8171229 40 143. USP51 ENST00000330856 8173174 40 144. HLA-DQA1 /// ENST00000383127 8178193 40 HLA-DRA 145. FAAH ENST00000243167 7901229 35 146. GDAP2 ENST00000369443 7918955 35 147. CD48 ENST00000368046 7921667 35 148. PTPRJ ENST00000278456 7939839 35 149. EXPH5 ENST00000265843 7951545 35 150. RPS26 /// ENST00000393490 7956114 35 LOC728937 /// RPS26L /// hCG_2033311 151. ALDH2 ENST00000261733 7958784 35 152. CALM1 ENST00000356978 7976200 35 153. NOX5 /// SPESP1 ENST00000395421 7984488 35 154. RHBDL1 ENST00000352681 7992010 35 155. CYLD ENST00000311559 7995552 35 156. OSBPL1A ENST00000357041 8022572 35 157. GYPC ENST00000259254 8045009 35 158. RQCD1 ENST00000295701 8048340 35 159. RBM44 ENST00000316997 8049552 35 160. — ENST00000384680 8051862 35 161. C3orf58 ENST00000315691 8083223 35 162. MFSD1 ENST00000264266 8083656 35 163. HACL1 ENST00000321169 8085608 35 164. SATB1 ENST00000338745 8085716 35 165. USP4 ENST00000351842 8087380 35 166. — ENST00000410125 8089928 35 167. — ENST00000384055 8097445 35 168. IL7R ENST00000303115 8104901 35 169. — ENST00000364497 8117018 35 170. FAM135A ENST00000370479 8120552 35 171. CD164 ENST00000310786 8128716 35 172. DYNLT1 ENST00000367088 8130499 35 173. NRCAM ENST00000379027 8142270 35 174. ZNF596 ENST00000308811 8144230 35 175. — ENST00000332418 8170322 35 176. TCEAL3 /// TCEAL6 ENST00000372774 8174134 35 177. SNAPIN ENST00000368685 7905598 30 178. DENND2D ENST00000369752 7918487 30 179. SAMD8 ENST00000372690 7928516 30 180. LHPP ENST00000368842 7931204 30 181. SLC37A2 ENST00000298280 7944931 30 182. FLI1 /// EWSR1 ENST00000344954 7945132 30 183. OR9G4 ENST00000395180 7948157 30 184. LOC338799 ENST00000391388 7967210 30 185. HEXDC ENST00000337014 8010787 30 186. NOTUM ENST00000409678 8019334 30 187. MCOLN1 ENST00000394321 8025183 30 188. PRKACA ENST00000350356 8034762 30 189. CRIM1 ENST00000280527 8041447 30 190. CECR5 ENST00000336737 8074227 30 191. RNF13 ENST00000392894 8083310 30 192. 40969 ENST00000339875 8103508 30 193. ZNF366 ENST00000318442 8112584 30 194. — ENST00000410754 8120979 30 195. GIMAP5 ENST00000358647 8137257 30 196. — ENST00000362484 8147242 30 197. TFE3 ENST00000315869 8172520 30 198. RHOU ENST00000366691 7910387 25 199. MED8 ENST00000290663 7915516 25 200. CASQ2 ENST00000261448 7918878 25 201. NUDT5 ENST00000378940 7932069 25 202. C11orf73 ENST00000278483 7942932 25 203. PAK1 ENST00000356341 7950578 25 204. PRSS21 ENST00000005995 7992722 25 205. — ENST00000332418 7997907 25 206. BTBD12 ENST00000294008 7999008 25 207. DHRS13 ENST00000394901 8013804 25 208. CCDC102B ENST00000319445 8021685 25 209. BCL2 ENST00000398117 8023646 25 210. ZNF211 /// ZNF134 ENST00000396161 8031784 25 211. NDUFV2 ENST00000340013 8039068 25 212. MYCN ENST00000281043 8040419 25 213. — ENST00000385528 8045561 25 214. — ENST00000362957 8046522 25 215. CASP8 ENST00000264275 8047419 25 216. RTN4 ENST00000394611 8052204 25 217. PLCG1 ENST00000244007 8062623 25 218. MGC42105 ENST00000326035 8105146 25 219. EMB ENST00000303221 8112007 25 220. — ENST00000386433 8121249 25 221. COL21A1 ENST00000370817 8127201 25 222. LRP12 ENST00000276654 8152280 25 223. LMNA ENST00000368301 7906085 20 224. — ENST00000385567 7907535 20 225. — ENST00000362863 7926805 20 226. ZNF503 ENST00000372524 7934553 20 227. NLRX1 ENST00000397884 7944463 20 228. — ENST00000391173 7954775 20 229. NDRG2 ENST00000298687 7977621 20 230. TRAF7 ENST00000326181 7992529 20 231. KRT40 ENST00000400879 8015152 20 232. KRT40 ENST00000400879 8019604 20 233. DRD5 ENST00000304374 8053725 20 234. ZC3H8 ENST00000409573 8054664 20 235. MMP9 ENST00000372330 8063115 20 236. PLTP ENST00000372420 8066619 20 237. — ENST00000362686 8100476 20 238. SPEF2 ENST00000282469 8104856 20 239. LRRC16A ENST00000332168 8117243 20 240. FBXO9 AK095315 8120269 20 241. EEPD1 ENST00000242108 8132305 20 242. FCN1 ENST00000371807 8165011 20 243. EFNA3 ENST00000368408 7905918 15 244. — ENST00000314893 7910385 15 245. TMEM19 ENST00000266673 7957167 15 246. PLXNC1 ENST00000258526 7957570 15 247. NHLRC3 ENST00000379599 7968703 15 248. MBNL2 ENST00000397601 7969677 15 249. EIF5 ENST00000216554 7977058 15 250. PLEKHG4 ENST00000360461 7996516 15 251. COPS3 ENST00000268717 8013094 15 252. FAM171A2 ENST00000398346 8016033 15 253. LOC653653 /// AP1S2 ENST00000380291 8017210 15 254. VAPA ENST00000340541 8020129 15 255. MATK ENST00000395040 8032682 15 256. ACTR2 ENST00000377982 8042337 15 257. BPI ENST00000262865 8062444 15 258. ERG ENST00000398905 8070297 15 259. LAMB2 ENST00000305544 8087337 15 260. — BC090058 8133752 15 261. PHTF2 ENST00000248550 8133818 15 262. — ENST00000333261 8133902 15 263. C8orf55 ENST00000336138 8148559 15 264. PDE7A ENST00000379419 8151074 15 265. NAPRT1 ENST00000340490 8153430 15 266. HLA-DRA ENST00000383127 8179481 15 267. SLC22A15 ENST00000369503 7904226 10 268. FCGR1A /// ENST00000369384 7905047 10 FCGR1B /// FCGR1C 269. SLC27A3 ENST00000271857 7905664 10 270. ID3 ENST00000374561 7913655 10 271. TBCEL ENST00000284259 7944623 10 272. FAM138D ENST00000355746 7960172 10 273. POMP ENST00000380842 7968297 10 274. SNN ENST00000329565 7993259 10 275. MED13 ENST00000262436 8017312 10 276. ZFP36L2 ENST00000282388 8051814 10 277. UXS1 ENST00000409501 8054395 10 278. CD40 ENST00000279061 8063156 10 279. — ENST00000362620 8066960 10 280. GGT5 ENST00000327365 8074991 10 281. — BC035666 8103023 10 282. G6PD ENST00000393562 8176133 10 283. — ENST00000384272 7902365 5 284. CLCC1 ENST00000369971 7918255 5 285. SCGB2A1 ENST00000244930 7940626 5 286. GAA ENST00000302262 8010354 5 287. SERPINB2 ENST00000404622 8021635 5 288. GPI ENST00000356487 8027621 5 289. LASS6 ENST00000392687 8046086 5 290. EIF4A2 AB209021 8084704 5 291. HLA-DRA ENST00000383127 8118548 5 292. — ENST00000385586 8136889 5 293. ANXA2P2 M62898 /// 8154836 5 NR_003573 294. FANCG ENST00000378643 8160935 5 295. FAM53B ENST00000337318 7936884 0 296. RFXAP ENST00000255476 7968653 0 297. UBR1 ENST00000382177 7987981 0 298. TBC1D2B ENST00000409931 7990657 0 299. SERPINB10 ENST00000397996 8021645 0 300. SEC23B ENST00000377481 8061186 0 301. MN1 ENST00000302326 8075126 0 302. CRTAP ENST00000320954 8078450 0

[0251] List of potential predictor genes for respiratory chemical sensitization, identified by ANOVA and backward elimination. Genes are annotated with Entrez Gene ID where found (www.ncbi.nlm.nih.gov/gene). The Affymetrix Probe Set ID for the Human ST 1.0 Array are provided. The validation call frequency (%) is the occurrence of each gene in the 20 Validation Biomarker Signatures obtained during cross-validation.

TABLE-US-00002 TABLE 2 Concentrations and vehicles used for each reference chemical. Max solubility Rv90 GARD input Compound Abbreviation Vehicle (μM) (μM) concentration (μM) Respiratory sensitizers Ammonium hexachloroplatinate AH Water 35 — 35 Ammonium persulfate AP DMSO — — 500 Glutaraldehyde GA Water — 10 10 Hexamethylen diisocyanate HDI DMSO 100  — 100 Maleic Anhydride MA DMSO — — 500 Methylene diphenol diisocyanate MDI DMSO 50 — 50 Phtalic Anhydride PA DMSO 200  — 200 Toluendiisocyanate TDI DMSO 40 — 40 Trimellitic anhydride TMA DMSO 150  — 150 Non-sensitizers 1-Butanol BUT DMSO — — 500 4-Aminobenzoic acid PABA DMSO — — 500 Chlorobenzene CB DMSO 98 — 98 Dimethyl formamide DF Water — — 500 Ethyl vanillin EV DMSO — — 500 Isopropanol IP Water — — 500 Methyl salicylate MS DMSO — — 500 Propylene glycol PG Water — — 500 Potassium permanganate PP Water 38 — 38 Tween 80 T80 DMSO — — 500 Zinc sulphate ZS Water 126  — 126 List of concentrations and vehicles used for each reference chemical used for assay development. Reference chemicals were classified as respiratory sensitizers or non-respiratory sensitizers through clinical observations in humans.

TABLE-US-00003 TABLE 4 Canonical Pathways associated with GRPS. Canonical Pathway -log(p-value) Regulated molecules.sup.1 TREM1 Signaling 5.4 CASP1, CCL2, CCL3, CD40, CD86, FCGR2B, IL8, IL1B, MPO, PLCG1, SIGIRR, TLR1, TLR6 Altered T Cell and B Cell 3.7 CD40, CD86, CD79A, FAS, FCER1G, HLA- Signaling in Rheumatoid DQA1, HLA-DRA, IL1B, IL1RN, PRTN3, SPP1, Arthritis TLR1, TLR6 Nicotinate and Nicotinamide 3.6 CD38, CDK6, DFFB, ENPP2, GRK5, MAP2K1, Metabolism MAPK6, NADK, NAPRT1, NNT, PAK1, PPM1F, PTPRJ, PTPRO, SGK1 Communication between 2.9 CCL3, CD40, CD86, FCER1G, HLA-DRA, IFNA5, Adaptive and Innate Immune IL8, IL1B, IL1RN, TLR1, TLR6 Cells B Cell Development 2.9 CD40, CD86, CD79A, HLA-DQA1, HLA-DRA, IL7R Sphingolipid Metabolism 2.6 ASAH2, CERK, CERS6, FUT4, KDSR, NAAA, PPM1F, PTPRJ, PTPRO, SPHK2, SPTLC2 Cell Cycle Control of 2.6 CDK6, CDT1, MCM2, MCM4, MCM6, MCM7 Chromosomal Replication Riboflavin Metabolism 2.6 ACPP, ENPP2, PPM1F, PTPRJ, PTPRO Glutathione Metabolism 2.5 G6PD, GGT5, GGTLC2, GLRX, GSTA3, H6PD, IDH2, MGST1, Aryl Hydrocarbon Receptor 2.4 AHR, ALDH1A1, CDK6, CDKN1A, CYP1B1, FAS, Signaling GSTA3, IL1B, JUN, MCM7, MGST1, NCOA3, NQO1, NQO2, RB1 Graft-versus-Host Defense 2.3 CD86, FAS, FCER1G, HLA-DQA1, HLA-DRA, Signaling IL1B, IL1RN Dendritic Cell Maturation 2.3 CD40, CD86, CD1A, CD1B, CD1C, CREB3L4, FCER1G, FCGR2A, FCGR2B, HLA-DQA1, HLA-DRA, IFNA5, IL1B, IL1RN, MAPK12, PIK3CD, PLCG1 CD28 Signaling in T-Helper 2.3 ACTR2, CALM1, CD86, FCER1G, HLA-DQA1, Cells HLA-DRA, JUN, MAP2K1, MAPK12, PAK1, PDPK1, PIK3CD, PLCG1 Lipid Antigen Presentation 2.3 CD1A, CD1B, CD1C, FCER1G by CD1 Cytotoxic T Cell Mediated 2.2 BCL2, CASP8, DFFB, FAS, FCER1G, HLA- Apoptosis of Target Cells DQA1, HLA-DRA Fatty Acid Biosynthesis 2.1 ACACA, FASN, SLC27A3 Autoimmune Thyroid 2.0 CD40, CD86, FAS, FCER1G, HLA-DQA1, HLA-DRA Disease Signaling .sup.1Molecules indicated in bold are present in the GARD Respiratory Prediction Signature. Molecules colored red are up

[0252] Table 4 Legend. Top Canonical Pathways associated with the top 1029 predictors able to separate respiratory chemical sensitizers from non-sensitizers. Molecules indicated in bold are present in the GRPS. Molecules colored red are up regulated in chemical respiratory sensitizers, while molecules colored green are down regulated in chemical respirator sensitizers.