BIOMARKER AND DIAGNOSIS SYSTEM FOR COLORECTAL CANCER DETECTION

Abstract

The present disclosure provides a biomarker for detecting colorectal cancer and a use thereof. A metabolomics method is used to analyze metabolites with significant differences in urine of patients with colorectal cancer and normal people, such that a series of biomarkers capable of early predicting an occurrence risk of colorectal cancer are screened out, a group of biomarkers are further screened to construct a diagnostic model for colorectal cancer, and the model can be used for conveniently, non-invasively and effectively predicting whether an individual suffers from colorectal cancer, and meets clinical needs.

Claims

1-20. (canceled)

21. A system for predicting whether an individual suffers from colorectal cancer, wherein the system comprises a data analysis module; and the data analysis module is configured to analyze a detection value of a biomarker, and the biomarker consists of 4-hydroxyphenylpyruvate, dimethylguanidinovaleric acid, N-methyl-4-aminobutyric acid, nicotinamide, p-cresol glucuronide, p-cresol sulfate, phenylacetylalanine, phenylacetylglutamine, phenylacetylmethionine, and phenylacetylthreonine.

22. The system according to claim 21, wherein the detection value of the biomarker is obtained by detecting the biomarker in a urine sample.

23. The system according to claim 22, wherein the detection value of the biomarker is obtained by detecting the presence or relative abundance or concentration of the biomarker in the urine sample of the individual.

24. The system according to claim 23, wherein the data analysis module adopts a random forest or a logistic regression equation to construct a model for analysis.

25. The system according to claim 24, wherein the data analysis module calculates a predictive value for predicting whether an individual suffers from colorectal cancer by substituting the detection value of the biomarker into the logistic regression equation to evaluate whether the individual suffers from the colorectal cancer.

26. The system according to claim 25, wherein the logistic regression equation is:
Z=4-hydroxyphenylpyruvate*0.037986+dimethylguanidinovaleric acid*0.4818−N-methyl-4-aminobutyric acid*1.0077−nicotinamide*1.525−p-cresol glucuronide*0.0353−p-cresol sulfate*0.021798−phenylacetylalanine*0.1902+phenylacetylglutamine*0.858−phenylacetylmethionine*0.118805+phenylacetylthreonine*0.59727+0.7486, $p=\frac{1}{1+e^z}$ wherein e is the base of the natural logarithm; and p is the predictive value for predicting whether the individual suffers from the colorectal cancer.

27. The system according to claim 26, wherein when p is greater than 0.5, the individual is predicted to have a high probability of colorectal cancer; and when p is less than 0.5, the individual is predicted to have a low probability of colorectal cancer.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0054] FIG. 1 is the flow chart of screening biomarkers in urine by metabolomics in example 1;

[0055] FIG. 2 shows the structural formula of 3-indoxyl sulfate in example 1;

[0056] FIG. 3 shows the structural formula of 4-hydroxyphenylacetylglutamine in example 1;

[0057] FIG. 4 shows the structural formula of 5-hydroxyindole glucuronide in example 1;

[0058] FIG. 5 shows the structural formula of phenylacetylglutamate in example 1;

[0059] FIG. 6 shows the structural formula of phenylacetylhistidine in example 1;

[0060] FIG. 7 shows the structural formula of phenylacetylmethionine in example 1;

[0061] FIG. 8 shows the structural formula of phenylacetylthreonine in example 1;

[0062] FIG. 9 is the schematic diagram of comparison of prediction accuracy of a colorectal cancer diagnostic model constructed by selecting 2, 3, 5, 10, 20, and 26 biomarkers respectively from 26 biomarkers in example 2;

[0063] FIG. 10 shows an ROC curve of a random forest model for predicting colorectal cancer constructed in example 2;

[0064] FIG. 11 is an analysis map of the random forest model for predicting colorectal cancer in example 2;

[0065] FIG. 12 an ROC curve of a logistic regression model for predicting colorectal cancer constructed in example 2;

[0066] FIG. 13 is an analysis map of the logistic regression model for predicting colorectal cancer in example 2; and

[0067] FIG. 14 shows an accuracy evaluation result of a colorectal cancer model in example 3.

DETAILED DESCRIPTION OF THE INVENTION

[0068] The present disclosure is further described in detail below with reference to the accompanying drawings and examples. It should be pointed out that the following examples are intended to facilitate the understanding of the present disclosure without any limitation. The reagents used in the examples are known and commercially available products.

Example 1 Screening Biomarkers of Colorectal Cancer in Urine by Metabolomics

[0069] In the example, through a non-targeted metabolomics research, urine samples of a healthy group and a colorectal cancer patient group were analyzed by using an ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). Besides, metabolites with significant differences between a colorectal cancer sample and a control sample were respectively screened by using four statistical methods of random forest, PLS-DA, volcano, and SVM. The screened metabolites with significant differences in the four statistical analysis methods were selected, finally 26 urine metabolites were obtained and used as biomarkers, and the functions of the biomarkers in the diagnosis or distinguishment of colorectal cancer were verified (see FIG. 1 for the flow chart).

[0070] Specific steps were as follows:

[0071] 1. Experimental Method

[0072] (1) Sample Collection

[0073] Urine samples were collected from 50 patients with colorectal cancer and 50 control individuals (non-colorectal cancer individuals). The patients with colorectal cancer were individuals with colorectal cancer confirmed by a colonoscopy.

[0074] (2) Sample Treatment

[0075] Methanol was added into the urine samples in a proportion of 1:4, the urine samples were shaken for 3 min to be mixed well, and the mixture was centrifuged at 20° C. and 4,000×g for 10 min. 100 μL of supernatant of each of 4 samples was put into 4 sample plates and blow-dried with nitrogen, and a complex solution was added for a subsequent LC-MS/MS detection.

[0076] (3) LC-MS/MS Detection and Data Processing

[0077] m/z ions were extracted from original mass spectrometry data detected by LC-MS/MS, a database was searched to retrieve and identify metabolites, chromatographic peak integrals of the metabolites were examined to obtain peak areas, data normalization and missing value filling were performed to obtain a data matrix to perform subsequent bioinformatic analysis, including four statistical methods of random forest, PLS-DA (partial least squares), volcano (volcano plot), and SVM (support vector machine), and the most effective differential metabolite ranking lists for sample grouping were respectively screened between the colorectal cancer samples and the control samples. Finally, the metabolites screened in the four methods were selected as biomarkers for colorectal cancer.

[0078] 2. Experimental Results

[0079] 32, 41, 35, and 52 different metabolites were screened by four statistical methods of random forest, PLS-DA, difference test and SVM, wherein 26 metabolites, i.e. 26 biomarkers, were screened in the four data analysis methods, as shown in Table 1.

TABLE-US-00001 TABLE 1 25 biomarkers for colorectal cancer Serial No. English Name CAS code Molecular formula  1 2-piperidinone 675-20-7 C.sub.5H.sub.9NO  2 3-hydroxyanthranilate 548-93-6 C.sub.7H.sub.7NO.sub.3  3 3-indoxyl sulfate — C.sub.8H.sub.7NO.sub.4S (structural formula shown in FIG. 2)  4 4-hydroxyphenyl- — C.sub.13H.sub.15NO.sub.6 acetylglutamine (structural formula shown in FIG. 3)  5 4-hydroxyphenylpyruvate 156-39-8 C.sub.9H.sub.8O.sub.4  6 5-hydroxyindole glucuronide — C.sub.8H.sub.7NOC.sub.6H.sub.8O.sub.6 (structural formula shown in FIG. 4)  7 6-hydroxyindole sulfate 487-94-5 C.sub.8H.sub.7NO.sub.4S  8 dimethylguanidinovaleric acid 107347-90-0 C.sub.8H.sub.15N.sub.3O.sub.3 (DMGV)  9 N-acetyl-cadaverine 32343-73-0 C.sub.7H.sub.16N.sub.2O 10 N-formylmethionine 4289-98-9 C.sub.6H.sub.11NO.sub.3S 11 nicotinamide 98-92-0 C.sub.6H.sub.6N.sub.2O 12 nicotinamide N-oxide 1986-81-8 C.sub.6H.sub.6N.sub.2O.sub.2 13 N-methyl-GABA 1119-48-8 C.sub.5H.sub.11NO.sub.2 14 p-cresol glucuronide 17680-99-8 C.sub.13H.sub.16O.sub.7 15 p-cresol sulfate 3233-58-7 C.sub.7H.sub.8O.sub.4S 16 phenylacetylalanine 17966-65-3 C.sub.11H.sub.13NO.sub.3 17 phenylacetylglutamate — C.sub.13H.sub.15NO.sub.5 (structural formula shown in FIG. 5) 18 phenylacetylglutamine 28047-15-6 C.sub.13H.sub.16N.sub.2O.sub.4 19 phenylacetylhistidine — C.sub.6H.sub.9N.sub.3O.sub.2C.sub.8H.sub.6O (structural formula shown in FIG. 6) 20 phenylacetylmethionine — C.sub.5H.sub.11NO.sub.2SC.sub.8H.sub.6O (structural formula shown in FIG. 7) 21 phenylacetylserine 65445-69-4 C.sub.11H.sub.13NO.sub.4 22 phenylacetyltaurine 33953-90-1 C.sub.10H.sub.13NO.sub.4S 23 phenylacetylthreonine — C.sub.4H.sub.9NO.sub.3C.sub.8H.sub.6O (structural formula shown in FIG. 8) 24 trimethylamine N-oxide 1184-78-7 C.sub.3H.sub.9NO 25 xanthine 69-89-6 C.sub.5H.sub.4N.sub.4O.sub.2 26 trizma acetate 6850-28-8 C.sub.6H.sub.15NO.sub.5

Example 2 Prediction Model for Colorectal Cancer

[0080] In the example, single biomarkers or a combination of multiple biomarkers screened in example 1 were used to establish prediction or diagnosis models for colorectal cancer. These models were used to distinguish colorectal cancer from non-colorectal cancer, to screen a patient with colorectal cancer from the population, or to predict whether an individual is a patient with colorectal cancer or the possibility of an individual suffering from colorectal cancer. Specific models were as follows.

[0081] 1. Single Biomarkers

[0082] An R language software was used to process data. According to the grouping of patients with colorectal cancer and a non-colorectal cancer population, the concentration changes of 26 biomarkers in the urine samples of the patients with colorectal cancer and the non-colorectal cancer population were determined. All the detection results were subjected to an LASSO regression analysis to establish a mathematical model to predict whether an individual suffers from colorectal cancer, and the effectiveness of the regression model was evaluated by using a calibration curve and an ROC curve.

[0083] The analysis results showed that 26 biomarkers were significantly correlated with colorectal cancer. The analysis results were shown in Table 2 and Table 3.

TABLE-US-00002 TABLE 2 Comparison of correlation detection results of 26 biomarkers and colorectal cancer 95% CI Indexes β OR p-value Lower Upper 2-piperidinone −5.302796177 0.004977656 0.0021997025 −2.439395904 −0.554916096 3-hydroxyanthranilate #N/A #N/A 0.000195065 −1.607767298 −0.526480702 3-indoxyl sulfate #N/A #N/A 0.123231485 −1.158547932 0.140887932 4-hydroxyphenylacetyl- 0.037986131 1.038716827 0.036132216 −3.772580625 −0.128923375 glutamine 4-hydroxyphenylpyruvate #N/A #N/A 0.036132216 −3.772580625 −0.128923375 5-hydroxyindole #N/A #N/A 0.19451781 −1.28588006 0.26590806 glucuronide 6-hydroxyindole #N/A #N/A 0.214792551 −1.294562772 0.294750772 sulfate dimethylguanidino 0.481847118 1.619062241 0.002751023 −3.214983555 −0.704188445 valeric acid N-acetyl-cadaverine #N/A #N/A 0.006027911 −5.694234145 −1.003437854 N-formylmethionine #N/A #N/A 0.006200771 −1.193531098 −0.204336902 nicotinamide −1.525090436 0.217601377 0.011443056 0.132060748 1.012671252 nicotinamide N-oxide #N/A #N/A 0.0000369156 0.543634687 1.434193313 N-methyl-4- −1.007770314 0.365031979 0.000151406 0.380868329 1.147055671 aminobutyric acid p-cresol glucuronide −0.035366893 0.965251207 0.005961446 −10.71959689 −1.875015108 p-cresol sulfate −0.021798367 0.978437501 0.004011742 −3.683079137 −0.724172863 phenylacetylalanine −0.190202421 0.826791757 0.021098845 −3.439994011 −0.286757989 phenylacetylglutamate #N/A #N/A 0.027185752 −2.7677993445 −0.170452655 phenylacetylglutamine 0.858050782 2.358558865 1.02818E−05 −1.93344453 −0.78233147 phenylacetylhistidine #N/A #N/A 0.0015908 −2.25387961 −0.54944039 phenylacetylmethionine −0.118805316 0.88798066 0.001919024 −3.178504783 −0.750631217 phenylacetylserine #N/A #N/A 0.00005738 −2.447360211 −0.890135789 phenylacetyltaurine #N/A #N/A 0.017478353 −2.948912433 −0.294979567 phenylacetylthreonine 0.597275285 1.817160804 0.002659366 −2.782042717 −0.609085283 trimethylamine N-oxide #N/A #N/A 0.00416445 −0.660183306 −0.127504694 xanthine #N/A #N/A 0.828967916 −0.657362597 0.818398597 trizma acetate #N/A #N/A 0.000041591 −111.298499 −42.38852896

TABLE-US-00003 TABLE 3 ROC analysis results of single biomarkers Serial AUC Cut- No. Biomarkers value Sensitivity Specificity off  1 2-piperidinone 0.7156 0.925  0.68 0.72  2 3-hydroxyanthranilate 0.7218  0.53175 0.52 0.82  3 3-indoxyl sulfate 0.7096 0.8711 0.62 0.76  4 4-hydroxy- 0.7036  1.24985 0.74 0.62 phenylacetylglutamine  5 4-hydroxyphenylpyruvate 0.7668  0.97835 0.72 0.76  6 5-hydroxyindole 0.7112 0.3662 0.46 0.96 glucuronide  7 6-hydroxyindole sulfate 0.6864  0.63085 0.48 0.86  8 dimethylguanidinovaleric 0.722  0.2471 0.58 0.82 acid  9 N-acetyl-cadaverine 0.7796 0.582  0.54 0.9  10 N-formylmethionine 0.6568  0.20645 0.28 0.98 11 nicotinamide 0.6324  2.27625 0.32 0.98 12 nicotinamide N-oxide 0.772  0.1686 0.88 0.58 13 N-methyl-4-aminobutyric 0.7444 1.1929 0.62 0.78 acid 14 p-cresol glucuronide 0.7836 0.86  0.64 0.64 15 p-cresol sulfate 0.7348 0.7536 0.64 0.82 16 phenylacetylalanine 0.7428 1.6654 0.8  0.58 17 phenylacetylglutamate 0.6988 1.0442 0.68 0.72 18 phenylacetylglutamine 0.7876 0.5643 0.62 0.84 19 phenylacetylhistidine 0.7478  0.96145 0.72 0.7  20 phenylacetylmethionine 0.7768  0.73925 0.7  0.78 21 phenylacetylserine 0.78  1.116  0.74 0.68 22 phenylacetyltaurine 0.6968 0.6231 0.5  0.84 23 phenylacetylthreonine 0.7352  1.21925 0.72 0.7  24 trimethylamine N-oxide 0.6708 0.9524 0.66 0.7  25 xanthine 0.774  0.8734 0.78 0.68 26 trizma acetate 0.7354 0.72  0.86 0.72

[0084] The correlation between the concentration changes of the 26 biomarkers and the colorectal cancer can be distinguished by OR values, p-values and the like in Table 2, and also can be distinguished by AUC values and the like in Table 3, wherein the OR values and the AUC values were most visual and obvious. The higher OR value indicated that the patients with colorectal cancer had a greater impact on the index compared with non-colorectal cancer patients, and the index exposure was more obvious. The higher AUC value indicated that the biomarker could more accurately distinguish between the colorectal cancer population and the non-colorectal cancer population.

[0085] It can be seen from Table 2 that the concentration changes of the 26 biomarkers were obviously correlated with colorectal cancer, wherein the phenylacetylglutamine had the highest correlation, with an OR value of 2.36, followed by phenylacetylthreonine, with an OR value of 1.82.

[0086] It can be seen from Table 3 that the AUC value of the concentration change of any of 26 biomarkers used alone to distinguish the colorectal cancer population and the non-colorectal cancer population can reach 0.63 or more, with high accuracy. The phenylacetylglutamine had the highest AUC value of 0.7876, followed by p-cresol glucuronide having the AUC value of 0.7836.

[0087] 2. Combination of Multiple Biomarkers

[0088] Although a single biomarker can also be used to distinguish urine samples of colorectal cancer from non-colorectal cancer or predict colorectal cancer. It is generally more accurate to combine multiple biomarkers for distinguishment or prediction.

[0089] However, the single biomarker with higher accuracy in predicting colorectal cancer does not necessarily play a larger role in the combination when combined with other one or more biomarkers. At the same time, the more number of the biomarkers does not indicate higher accuracy of prediction (AUC value) of the combination. Therefore, a large number of verification experiments are required.

[0090] Since the AUC and OR values of the biomarkers are biased toward evaluating the relative importance of the variables in the statistical models and are not suitable for constructing a model for the preferred variables, the example preferably used 2, 3, 5, 10, 20, and 26 biomarkers with the highest concentration fold change in the urine samples of colorectal cancer and non-colorectal cancer to construct a diagnostic model for colorectal cancer. The concentration fold change (fold change=expression mean value of disease sample divided by expression mean value of normal sample) of the 26 biomarkers in the urine samples of colorectal cancer and non-colorectal cancer ranked from high to low, and the results were shown in Table 4.

TABLE-US-00004 TABLE 4 Ranking of concentration fold changes of 26 biomarkers in urine samples of colorectal cancer and non-colorectal cancer Fold Rank Biomarkers Change T-tests AUC  1 p-cresol sulfate 5.0115 4.1426E−11 0.7348  2 phenylacetylthreonine 4.8447 6.0512E−10 0.7352  3 N-methyl-4-aminobutyric 3.0586 7.534E−7 0.7444 acid  4 4-hydroxyphenylpyruvate 2.8178 2.2337E−5  0.7668  5 phenylacetylmethionine 2.7238 3.8753E−7  0.7768  6 p-cresol glucuronide 2.7028 1.7998E−7  0.7836  7 nicotinamide 2.0965 1.9958E−6  0.6324  8 phenylacetylalanine 2.0369 7.5589E−6  0.7428  9 phenylacetylglutamine 1.8305 4.9111E−8  0.7876 10 dimethylguanidinovaleric 1.8246 7.6888E−5  0.722  acid 11 3-hydroxyanthranilate 1.7392 8.2523E−4  0.7218 12 5-hydroxyindole 1.643  3.063E−4 0.7112 glucuronide 13 phenylacetylglutamate 1.6132 1.085E−7 0.6988 14 phenylacetylhistidine 1.5252 3.3237E−5  0.7478 15 2-piperidinone 1.4667 1.1365E−4  0.7156 16 N-formylmethionine 1.3568 4.7596E−4  0.6568 17 phenylacetyltaurine 1.2161 1.9028E−5  0.6968 18 3-indoxyl sulfate  0.98732 3.4274E−4  0.7096 19 6-hydroxyindole sulfate  0.92086 0.0019052 0.6864 20 trimethylamine N-oxide  0.77014 7.5794E−5  0.6708 21 4-hydroxyphenyl- −0.5916  0.34593 0.7036 acetylglutamine 22 N-acetyl-cadaverine −0.77292 0.18073 0.7796 23 trizma acetate −0.83338 0.0016428 0.7354 24 xanthine −1.0127  2.1818E−5  0.774  25 nicotinamide N-oxide −1.2215  0.003826 0.772  26 phenylacetylserine −1.7863  0.001003 0.78 

[0091] According to the concentration fold changes of the 26 biomarkers in the urine samples of colorectal cancer and non-colorectal cancer provided in Table 4, 2, 3, 5, 10, 20, and 26 biomarkers of the 26 biomarkers were selected respectively in the example to construct a diagnostic model of colorectal cancer through random forest.

[0092] The 2 biomarkers were the first and second biomarkers (p-cresol sulfate and phenylacetylthreonine) in Table 4. In the constructed random forest model, the information gain ratio (GINI coefficient) of the p-cresol sulfate was 25.31 and the mean decrease accuracy was 21.17; and the GINI coefficient of the phenylacetylthreonine was 24.22 and the mean decrease accuracy was 16.71.

[0093] The 3 biomarkers were the first to third biomarkers in Table 4. In the constructed random forest model, the GINI coefficient of the p-cresol sulfate was 15.43 and the mean decrease accuracy was 16.37; the GINI coefficient of the phenylacetylthreonine was 15.75 and the mean decrease accuracy was 15.04; and the GINI coefficient of the N-methyl-4-aminobutyric acid was 18.33 and the mean decrease accuracy was 24.42.

[0094] The 5 biomarkers were the first to fifth biomarkers in Table 4. In the constructed random forest model, the GINI coefficient of the p-cresol sulfate was 7.86 and the mean decrease accuracy was 10.99; the GINI coefficient of the phenylacetylthreonine was 6.39 and the mean decrease accuracy was 5.58; the GINI coefficient of the N-methyl-4-aminobutyric acid was 13.73 and the mean decrease accuracy was 25.36; the GINI coefficient of the 4-hydroxyphenylpyruvate was 10.43 and the mean decrease accuracy was 45.38; and the GINI coefficient of the phenylacetylmethionine was 11.05 and the mean decrease accuracy was 18.74.

[0095] The 10 biomarkers were the first to tenth biomarkers in Table 4. In the constructed random forest model, the GINI coefficient of the p-cresol sulfate was 3.64 and the mean decrease accuracy was 7.56; the GINI coefficient of the phenylacetylthreonine was 2.46 and the mean decrease accuracy was 4.80; the GINI coefficient of the N-methyl-4-aminobutyric acid was 8.04 and the mean decrease accuracy was 18.60; the GINI coefficient of the 4-hydroxyphenylpyruvate was 6.25 and the mean decrease accuracy was 12.60; the GINI coefficient of the phenylacetylmethionine was 6.26 and the mean decrease accuracy was 12.85; the GINI coefficient of the p-cresol glucuronide was 5.20 and the mean decrease accuracy was 11.07; the GINI coefficient of the nicotinamide was 6.56 and the mean decrease accuracy was 12.51; the GINI coefficient of the phenylacetylalanine was 3.18 and the mean decrease accuracy was 6.30; the GINI coefficient of the phenylacetylglutamine was 4.47 and the mean decrease accuracy was 6.83; and the GINI coefficient of the dimethylguanidinovaleric acid was 3.43 and the mean decrease accuracy was 9.16.

[0096] The 20 biomarkers were the first to twentieth biomarkers in Table 4. In the constructed random forest model, the GINI coefficient of the p-cresol sulfate was 2.36 and the mean decrease accuracy was 6.21; the GINI coefficient of the phenylacetylthreonine was 1.73 and the mean decrease accuracy was 4.02; the GINI coefficient of the N-methyl-4-aminobutyric acid was 5.92 and the mean decrease accuracy was 16.23; the GINI coefficient of the 4-hydroxyphenylpyruvate was 4.10 and the mean decrease accuracy was 9.28; the GINI coefficient of the phenylacetylmethionine was 3.79 and the mean decrease accuracy was 10.13; the GINI coefficient of the p-cresol glucuronide was 3.77 and the mean decrease accuracy was 9.49; the GINI coefficient of the nicotinamide was 4.67 and the mean decrease accuracy was 11.61; the GINI coefficient of the phenylacetylalanine was 2.26 and the mean decrease accuracy was 5.84; the GINI coefficient of the phenylacetylglutamine was 2.67 and the mean decrease accuracy was 7.71; the GINI coefficient of the dimethylguanidinovaleric acid was 2.00 and the mean decrease accuracy was 7.77; the GINI coefficient of the 3-hydroxyanthranilate was 2.03 and the mean decrease accuracy was 4.32; the GINI coefficient of the 5-hydroxyindole glucuronide was 2.69 and the mean decrease accuracy was 5.66; the GINI coefficient of the phenylacetylglutamate was 1.59 and the mean decrease accuracy was 4.38; the GINI coefficient of the phenylacetylhistidine was 1.62 and the mean decrease accuracy was 4.96; the GINI coefficient of the 2-piperidinone was 1.57 and the mean decrease accuracy was 1.85; the GINI coefficient of the N-formylmethionine was 1.45 and the mean decrease accuracy was 2.81; the GINI coefficient of the phenylacetyltaurine was 1.28 and the mean decrease accuracy was 0.79; the GINI coefficient of the 3-indoxyl sulfate was 1.41 and the mean decrease accuracy was 3.51; the GINI coefficient of the 6-hydroxyindole sulfate was 1.57 and the mean decrease accuracy was 1.93; and the GINI coefficient of the trimethylamine N-oxide was 1.02 and the mean decrease accuracy was 2.61.

[0097] The 26 biomarkers were the first to twenty-sixth biomarkers in Table 4. In the constructed random forest model, the GINI coefficient of the p-cresol sulfate was 1.69 and the mean decrease accuracy was 7.04; the GINI coefficient of the phenylacetylthreonine was 1.04 and the mean decrease accuracy was 2.80; the GINI coefficient of the N-methyl-4-aminobutyric acid was 3.57 and the mean decrease accuracy was 12.93; the GINI coefficient of the 4-hydroxyphenylpyruvate was 2.45 and the mean decrease accuracy was 5.50; the GINI coefficient of the phenylacetylmethionine was 2.68 and the mean decrease accuracy was 7.68; the GINI coefficient of the p-cresol glucuronide was 2.61 and the mean decrease accuracy was 8.31; the GINI coefficient of the nicotinamide was 2.56 and the mean decrease accuracy was 8.02; the GINI coefficient of the phenylacetylalanine was 1.47 and the mean decrease accuracy was 4.84; the GINI coefficient of the phenylacetylglutamine was 1.83 and the mean decrease accuracy was 5.74; the GINI coefficient of the dimethylguanidinovaleric acid was 1.34 and the mean decrease accuracy was 3.76; the GINI coefficient of the 3-hydroxyanthranilate was 1.14 and the mean decrease accuracy was 4.11; the GINI coefficient of the 5-hydroxyindole glucuronide was 1.76 and the mean decrease accuracy was 4.39; the GINI coefficient of the phenylacetylglutamate was 0.88 and the mean decrease accuracy was 3.11; the GINI coefficient of the phenylacetylhistidine was 1.00 and the mean decrease accuracy was 4.79; the GINI coefficient of the 2-piperidinone was 1.20 and the mean decrease accuracy was 1.80; the GINI coefficient of the N-formylmethionine was 0.79 and the mean decrease accuracy was 2.15; the GINI coefficient of the phenylacetyltaurine was 0.58 and the mean decrease accuracy was 2.70; the GINI coefficient of the 3-indoxyl sulfate was 0.96 and the mean decrease accuracy was 3.64; the GINI coefficient of the 6-hydroxyindole sulfate was 0.73 and the mean decrease accuracy was 2.70; the GINI coefficient of the trimethylamine N-oxide was 0.74 and the mean decrease accuracy was 2.33; the GINI coefficient of the 4-hydroxyphenylacetylglutamine was 0.83 and the mean decrease accuracy was 4.61; the GINI coefficient of the N-acetyl-cadaverine was 2.22 and the mean decrease accuracy was 7.72; the GINI coefficient of the trizma acetate was 2.48 and the mean decrease accuracy was 8.06; the GINI coefficient of the xanthine was 2.70 and the mean decrease accuracy was 8.67; the GINI coefficient of the nicotinamide N-oxide was 8.21 and the mean decrease accuracy was 16.94; and the GINI coefficient of the phenylacetylserine was 2.01 and the mean decrease accuracy was 7.16.

[0098] The AUC value and 95% confidence interval (CI) of the six random forest diagnostic models constructed with the above 2, 3, 5, 10, 20, and 26 biomarkers were calculated respectively, and the results were shown in FIG. 9.

[0099] It can be seen from FIG. 9 that the AUC value of the model constructed by selecting two biomarkers with the highest ranking among the 26 biomarkers can only reach 0.922, and the 95% CI was 0.718-0.999. As the number of the selected biomarkers increased, the AUC value gradually increased, and the 95% CI gradually decreased. When 10 biomarkers were selected to construct a diagnostic model for colorectal cancer, the AUC value reached 0.935 and the 95% CI was 0.842-0.998. However, when the number of the biomarkers further rose to 20 or 26, the space for AUC to continue to rise was very limited, and the confidence interval became larger. In addition, compared with 20 and 26 biomarkers, the use of 10 biomarkers to construct a model can reduce the number of variables and reduce the complexity of the model. Therefore, it is preferred to use the top 10 biomarkers in Table 4 to construct the diagnostic model for colorectal cancer, and thus very good prediction accuracy can be achieved and the model is simpler and more convenient.

[0100] 42 clinically known patients with colorectal cancer and 42 non-colorectal cancer patients were taken as the total data set to detect the biomarker detection values of the urine samples. The analysis was performed through the random forest model of 10 biomarkers. The analysis map was shown in FIG. 11. It can be seen from FIG. 11 that when the random forest model constructed with the 10 biomarkers was used to predict colorectal cancer, there would be some errors (of course, the errors were unavoidable). Among the 42 patients with colorectal cancer, 37 cases were detected. Among 42 non-colorectal cancer patients, 5 cases were classified as patients with colorectal cancer. The accuracy rate was 88%. It can be seen from FIG. 11, when a predictive value p was greater than 0.5, an individual was predicted to have a high probability of colorectal cancer; and when a predictive value p was less than 0.5, an individual was predicted to have a low probability of colorectal cancer.

[0101] The 10 biomarkers of the top 10 biomarkers of fold change were used for multivariate regression analysis to establish a logistic regression evaluation model to predict whether an individual suffered from colorectal cancer:

Z=4-hydroxyphenylpyruvate*0.037986+dimethylguanidinovaleric acid*0.4818−N-methyl-4-aminobutyric acid*1.0077−nicotinamide*1.525−p-cresol glucuronide*0.0353−p-cresol sulfate*0.021798−phenylacetylalanine*0.1902+phenylacetylglutamine*0.858−phenylacetylmethionine*0.118805+phenylacetylthreonine*0.59727+0.7486,

[00003]$p=\frac{1}{1+e^z}$ [0102] wherein e is the base of the natural logarithm; and p is a predictive value for predicting whether an individual suffers from colorectal cancer and the name of the biomarker represents the relative abundance of the corresponding biomarker in a urine sample, that is, a peak area of the biomarker in a detection spectrum obtained by an ultra-performance liquid chromatography-tandem mass spectrometry.

[0103] The ROC curve of the logistic regression model to predict whether an individual suffers from colorectal cancer provided in the example was shown in FIG. 12. The AUC value reached 0.957 and was significantly higher than that of the random forest model of 10 biomarkers.

[0104] The logistic regression model was used to predict whether an individual suffered from colorectal cancer. 50 clinically known patients with colorectal cancer and 50 non-colorectal cancer patients were taken as the total data set for analysis. The analysis results were shown in FIG. 13 and Table 5.

TABLE-US-00005 TABLE 5 Analysis results of model for predicting whether individual suffering from colorectal cancer Analysis results of logistic regression model Actual prediction Negative Positive Negative 46 4 Positive 5 45

[0105] It can be seen from FIG. 13 and Table 5 that the logistic regression evaluation model constructed by the 10 biomarkers to predict whether an individual suffered from colorectal cancer was used for analysis. Among 50 patients with colorectal cancer, 45 were detected. Among 50 non-colorectal cancer patients, 5 cases were classified as patients with colorectal cancer. The accuracy rate reached 90% or more, and thus was improved.

[0106] It can be seen from FIG. 13, p of 0.5 can be used as a dividing point for determination. When a predictive value p was greater than 0.5, an individual was predicted to have a high probability of colorectal cancer; and when a predictive value p was less than 0.5, an individual was predicted to have a low probability of colorectal cancer.

Example 3 Evaluation of Model for Predicting Colorectal Cancer

[0107] In the example, the accuracy of clinical application of the model for predicting colorectal cancer constructed in example 2 was evaluated. The above 42 patients with colorectal cancer and 42 non-colorectal cancer patients were taken as the total data set, from which 8 patients with CRC and 8 normal people (non-CRC patients) were randomly selected, and urine samples were taken. The relative abundance of the 10 biomarkers in the model was measured according to the sample processing method in example 1, so as to calculate the predictive value p through the model and predict whether an individual suffers from colorectal cancer. The results were shown in FIG. 14.

[0108] It can be seen from FIG. 14 that all the 8 patients with colorectal cancer were detected, and one of the 8 normal people was predicted to suffer from colorectal cancer, with an accuracy rate of 93.75%.

[0109] All the patents and publications mentioned in the description of the present disclosure indicate that these are public technologies in the art and can be used by the present disclosure. All the patents and publications cited herein are listed in the references, just as each publication is specifically referenced separately. The present disclosure described herein can be realized in the absence of any one element or multiple elements, one restriction or multiple restrictions, where the limitation is not specifically described here. For example, in each example, the terms “comprise”, “substantially composed of” and “composed of” can be replaced by the remaining two terms of either. The so-called “a” here only means “a kind”, not excluding only one, but also can indicate two or more. The terms and expressions used herein are descriptive, without limitation. Besides, there is no intention to indicate that these terms and interpretations described in the description exclude any equivalent features. However, it can be known that any appropriate changes or modifications can be made within the scope of the present disclosure and claims. It can be understood that the examples described in the present disclosure are some preferred examples and features. A person skilled in the art can make some modifications and changes according to the essence of the description of the present disclosure. These modifications and changes are also considered to fall within the scope of the present disclosure and the scope limited by independent claims and dependent claims.