METHOD FOR SORTING COLORECTAL CANCER AND ADVANCED ADENOMA AND USE OF THE SAME

Abstract

The present invention relates to a detecting method for colorectal cancer and advanced adenoma group, comprising measuring the relative expression level of MKi67, KRT19, EpCAM, TYMS, PPARG, MCAM, ANKHD1-EIF4EBP3, SNAI2, MMP23B, FOXA2, NPTN, GPR15, TERT, VIM, and ERBB2 genes or proteins encoded by the genes in sample, wherein if the MKi67, KRT19 and EpCAM genes are expressed higher than other genes, it is judged as a normal group, if the TYMS, PPARG, MCAM, and ANKHD1-EIF4EBP3 genes are expressed higher than other genes, it is judged as a colorectal cancer group, if the SNAI2, MMP23B, and FOXA2 genes are expressed higher than other genes, it is judged as an advanced adenoma group or colorectal cancer group, if the NPTN, GPR15, TERT, VIM and ERBB2 genes are expressed higher than other genes, it is judged as an advanced adenoma group.

Claims

1. A selectively detecting method for colorectal cancer and advanced adenoma group, comprising measuring the relative expression level of MKi67, KRT19, EpCAM, TYMS, PPARG, MCAM, ANKHD1-EIF4EBP3, SNAI2, MMP23B, FOXA2, NPTN, GPR15, TERT, VIM, and ERBB2 genes or proteins encoded by the genes in sample, wherein if the MKi67, KRT19 and EpCAM genes or proteins encoded by the genes are expressed higher than other genes or proteins encoded by those genes, it is judged as a normal group, if the TYMS, PPARG, MCAM, and ANKHD1-EIF4EBP3 genes or the proteins encoded by the genes are expressed higher than other genes or proteins encoded by those genes, it is judged as a colorectal cancer group, if the SNAI2, MMP23B, and FOXA2 genes or the proteins encoded by the genes are expressed higher than other genes or proteins encoded by those genes, it is judged as an advanced adenoma group or colorectal cancer group, if the NPTN, GPR15, TERT, VIM and ERBB2 genes or the proteins encoded by the genes are expressed higher than other genes or the protein encoded by those genes, it is judged as an advanced adenoma group.

2. The selectively detecting method according to claim 1, wherein the method for measuring the expression of the gene or the protein encoded by the gene is preferably characterized by measuring using primer and probe or using antibody.

3. The selectively detecting method according to claim 2, wherein the primer and probe comprise the sequences set forth in SEQ ID NOs: 1 to 46.

4. A kit for diagnosing colorectal cancer comprising a substance capable of measuring the relative expression levels of TYMS, PPARG, MCAM, and ANKHD1-EIF4EBP3 genes or proteins encoded by the genes.

5. The kit according to claim 4, wherein the substance capable of measuring the relative expression level of the gene is a primer and probe set.

6. The kit according to claim 5, wherein the primer and probe set consists of the sequences set forth in SEQ ID NOs: 1 to 3, SEQ ID NOs: 14 to 16, SEQ ID NOs: 17 to 19, and SEQ ID NOs: 26 to 28.

7. A kit for selectively detecting colorectal cancer and advanced adenomas, comprising a substance capable of measuring the relative expression level of proteins encoded by MKi67, KRT19 and EpCAM genes or proteins encoded by the genes, a substance capable of measuring the relative expression level of TYMS, PPARG, MCAM and ANKHD1-EIF4EBP3 genes or proteins encoded by the genes, a substance capable of measuring the relative expression level of SNAI2, MMP23B, and FOXA2 genes or proteins encoded by the genes, and a substance capable of measuring the relative expression levels of NPTN, GPR15, TERT, VIM, and ERBB2 genes or proteins encoded by the genes.

8. The kit according to claim 7, wherein the substance capable of measuring the relative expression level of the gene is a primer and probe set.

9. The kit according to claim 8, wherein the primer and probe set preferably consists of the sequences set forth in SEQ ID NOs: 1 to 46.

Description

DESCRIPTION OF DRAWINGS

[0073] FIG. 1 is a diagram showing a heatmap showing the expression pattern of each group of genes,

[0074] FIG. 2 is a diagram showing an overview of model construction and performance confirmation of an embodiment of the present invention;

[0075] FIG. 3 is a diagram showing the ROC Curve and PR Curve in the test set and

[0076] FIG. 4 is a diagram showing an outline of model construction and performance confirmation of a comparative example of the present invention.

MODE FOR INVENTION

[0077] Hereinafter, the present invention will be described in more detail by the following examples. However, the following examples are described with the intention of illustrating the present invention, and the scope of the present invention is not to be construed as being limited by the following examples.

Example 1; Collection of Clinical Specimens

[0078] From 2017 to 2022, Blood samples from subjects scheduled for colonoscopy were collected at the Shinchon Severance Hospital (Approval No. 4-2017-0148), the Gangnam Severance Hospital (Approval No. 3-2017-0024), the Kangbuk Samsung Hospital (Approval No. 2017-02-022-009) in the Department of Gastroenterology, the Health Examination Center of Wonju Severance Christian Hospital (approval number CR319115) with the approval of the Bioethics Review Board (IRB) of each institution. A total of 3 ml of blood was collected using a Tempus blood tube (Applied Biosystems®). Subjects were classified as follows through the results of colonoscopy (Table 1)

TABLE-US-00001 TABLE 1 No. of samples Classification Classification criteria (persons) Colorectal As a result of colonoscopy, 148 cancer subjects with cancer in the colon group Advanced As a result of colonoscopy, 289 adenoma subjects with advanced adenoma group in the colon Normal Subjects with no lesions in the 142 group large intestine as a result of colonoscopy Total 579
Table 1 shows the classification of subjects and the number of samples according to colonoscopy results.

Example 2: Isolation of Total RNA from Blood Specimens

[0079] Total RNA is isolated from a blood sample collected with a Tempus tube using the Tempus blood RNA isolation kit (Applied Biosystems®).

Example 3: cDNA Construction from Isolated Total RNA and qPCR

[0080] i. Complementary DNA (cDNA) Synthesis

[0081] Isolated total RNA 1.5˜4.5 ug, Random primer (3 ug/uL) (Invitrogen) 2.5 uL, dNTP mixture (2.5 mM each) (Intron) 2.5 uL, M-MLV reverse transcription polymerase (200 U/uL) (Invitrogen) 2.5 uL, 10 μL of 5× First-strand buffer (250 mM Tris-HCl) (Invitrogen), and 5 μL of Dithiothreitol (0.1 M) (Invitrogen) were added, and ultrapure water was added to a final volume of 50 μL, and mixed well. The synthetic reaction solution was reacted in a thermocycler (Applied Biosystems) at 25° C., 30 minutes—37° C., 50 minutes—70° C., 15 minutes to synthesize cDNA.

ii. Perform quantitative polymerase chain reaction (qPCR)

[0082] For the composition of the qPCR reaction, added 10 μL of THUNDERBIRD®Probe qPCR Mix (TOYOBO), Forward/Reverse Primer, Probe (10 pmole/uL) 1 μL, and added 2 μL of synthesized cDNA, and add ultrapure water to make the final volume 20 μL, and mixed. The qPCR reaction was performed using CFX96 (Biorad), and the reaction temperature conditions were as follows. After 95° C., 3 minutes, 95° C., 3 seconds—60° C., 30 seconds were repeated 40 times. Each time the annealing process (60° C., 30 seconds) was performed, a process of measuring fluorescence was added to measure the fluorescence value that increased by number of times. A constant fluorescence value was set as the threshold, and the Cq value, which is the number of cycles at the time of reaching the threshold, was derived.

Example 4: Confirmation of Results and Analysis of Relative Expression of Target Genes

[0083] Using the Cq value of the GAPDH gene used as an endogenous control, the relative expression level (2.sup.−ΔCq) of the target gene is calculated using the Cq value of the target gene. A list of targeted genes follows (Table 2).

2.sup.−ΔCq=2.sup.−(target gene Cq−GAPDH gene Cq)  [Calculation formula]

TABLE-US-00002 TABLE 2 No. Blood genetic markers 1 ANKHD1- ANKHD1-EIF4EBP3 Readthrough EIF4EBP3 2 EpCAM Epithelial Cell Adhesion Molecule 3 ERBB2 Erb-B2 Receptor Tyrosine Kinase 2 4 FOXA2 Forkhead Box A2 5 GPR15 G Protein-Coupled Receptor 15 6 KRT19 Keratin 19 7 MCAM Melanoma Cell Adhesion Molecule 8 MKi67 Marker Of Proliferation Ki-67 9 MMP23B Matrix Metallopeptidase 23B 10 NPTN Neuroplastin 11 PPARG Peroxisome Proliferator Activated Receptor Gamma 12 SNAI2 Snail Family Transcriptional Repressor 2 13 TERT Telomerase Reverse Transcriptase 14 TYMS Thymidylate Synthetase 15 VIM Vimentin
Table 2 is a list of target blood genetic markers

[0084] In order to compare the relative expression amount of each gene group, a heatmap based on the average relative expression amount of each gene group was constructed using the pheatmap package (version 1.0.12) of Statistical R software (version 3.6.3) (FIG. 1). When building a heatmap, colors are displayed according to the Z-score, and the Z-score calculation formula for each gene group is as follows. The lower the Z-score, the lower the expression compared to other groups, and the higher the Z-score, the higher the expression compared to other groups.

Z-score=(expression level of the group−average expression level in all groups)/(standard deviation between all groups)  [Calculation formula]

[0085] As a result, 3 genes (MKi67, KRT19, EpCAM) were highly expressed in the normal group compared to other groups and 4 genes (TYMS, PPARG, MCAM, ANKHD1-EIF4EBP3) were highly expressed in the colorectal cancer group compared to other groups, and 3 genes (SNAI2, MMP23B, FOXA2) were highly expressed in the advanced adenoma group and colorectal cancer group compared to other groups, and five genes (NPTN, GPR15, TERT, VIM, ERBB2) were highly expressed in the advanced adenoma group.

Example 5: Establishment of a Classification Model for the Purpose of Screening for Colorectal Cancer and Advanced Adenoma by Substituting the Relative Expression Level of Target Genes

[0086] An artificial intelligence algorithm-based classification model was constructed using the H2O package (version 3.32.1.3) of Statistical R software (version 3.6.3). The production of colorectal cancer and advanced adenoma diagnosis prediction models was based on Deep neural network (DNN), Generalized linear model (GLM), Random Forest (RF), and Gradient boosting machine (GBM) algorithms, and several types of models (GLM, RF, DNN, GBM, stacked ensemble (SE)) was performed by grafting Automated machine learning (AutoML) method to build a model suitable for data, but is not limited thereto.

[0087] By dividing the entire sample into a training set and a test set, and by substituting the results of the training set, an artificial intelligence algorithm-based classification model that can distinguish between the colorectal cancer group and the advanced glandular group compared to the normal group is constructed, and the performance of the built model is evaluated using the test set (FIG. 2).

[0088] When building a model using a training set, a 5-fold cross-validation technique is applied so that the training set is divided into 5 areas and so at the same time as learning the model, the performance of the model was verified using each area to build a high-performance model.

[0089] The performance of the artificial intelligence classification model was judged through the AUROC and AUPRC values of the training set and test set based on the AUROC and AUPRC values, which are representative performance indicators of the classification model. Among them, the model with the best performance was selected based on the performance of the new test set that was not used for model learning.

[0090] The AUROC and AUPRC values of the GLM, DNN, GBM, and RF models built based on each algorithm and the SE model built through AutoML are as follows (Table 3). As a result, the AUROC and AUPRC indicators were the highest in the SE model based on the test set (FIG. 3).

TABLE-US-00003 TABLE 3 Training set Test set Model AUROC AUPRC AUROC AUPRC GLM 0.91 0.97 0.87 0.96 RF 0.92 0.97 0.95 0.98 DNN 0.90 0.96 0.90 0.97 GBM 1.00 1.00 0.95 0.99 AutoML 1.00 1.00 0.97 0.99 (SE)

[0091] Table 3 shows AUROC and AUPRC performance indicators in the training set and test set.

[0092] As a result of confirming the sensitivity and specificity of each group in the SE model, as shown in Table 4, the sensitivity to classify the colorectal cancer group was 91.9%, the sensitivity to classify the advanced adenoma group was 92.6%, and the specificity to classify the normal group was 91.7%.

TABLE-US-00004 TABLE 4 Result of Test set (Total 154 persons) Positive Negative Sensitivity Specificity Classification (persons) (persons) (%) (%) Colorectal 110 8 92.4 cancer group + Advanced adenoma group(n = 118) colorectal 34 3 91.9 cancer group (n = 37) Advanced 75 6 92.6 adenoma group(n = 81) Normal group 3 33 91.7 (n = 36)
Table 4 shows the sensitivity and specificity results for each group of the SE model.

TABLE-US-00005 [  5] Primer Primer's and and Taqman probe's PCR Target TaqMan Sequence sequence product gene probe No (5′ --> 3′) (bp) PpARG Forward 1 CCC TTC ACT ACT GTT GAC 133 TTCTC Taqman 2 FAM-TCA CAA GAA CAG probe ATC CAG TGG TTG CA-BHQ1 Reverse 3 CTT TGA TTG CAC TTT GGT ACT CTT KRT19 Forward 4 GAT GAG CAG GTC CGA 96 GGT TA Taqman 5 FAM-CTG CGG CGC ACC probe CTT CAG GGT CT-BHQ1 Reverse 6 TCT TCC AAG GCA GCT TTC AT EPCAM Forward 7 GCC AGT GTA CTT CAG TTG 82 GTG CAC Taqman FAM-TAC TGT CAT TTG CTC probe 8 AAA GCT GGC TGC CA- BHQ1 Reverse 9 CAT TTC TGC CTT CAT CAC CAA ACA ERBB2 Forward 10 AAG CAT ACG TGA TGG 115 CTG GTG T Taqman 11 FAM-ATA TGT CTC CCG CCT probe1 TCT GGG CAT CT-BHQ1 Taqman 12 FAM-CAT CCA CGG TGC probe2 AGC TGG TGA CAC A-BHQ1 Reverse 13 TCT AAG AGG CAG CCA TAG GGC ATA MCAM Forward 14 TTC TGA AGT GCG GCC TCT 74 CC Taqman 15 FAM-TCC CAA GGC AAC probe CTC AGC CAT GTC G-BHQ1 Reverse 16 CGC TTC TCC TTG TGG ACA GAA AAC ANKHD1- Forward 17 TTCAGTCCCTGCTCTCAAA 108 EIF4EBP3 Taqman 18 FAM- probe ACCGAAGAAGAGAATTGG ACGGCC-BHQ1 Reverse 19 ATCCTGGTGCCTCTGGTTA GPR15 Forward 20 CTG TGT CAA CCC TTT CAT 106 TTAC Taqman 21 FAM-CAT TGT CCA CTG CTT probe GTG CCC TTG-BHQ1 Reverse 22 GTG CTA CTC CCA AAG TCA TAG MMP23B Forward 23 ACC TCC GGA TAG GCT TCT 136 A Taqman 24 FAM- probe ATCAACCACACGGACTGCC TGG-BHQ1 Reverse 25 CTG TCG TCG AAG TGG ATG C TYMS Forward 26 CTGAAGCCAGGTGACTTTA 90 TAC Taqman 27 FAM- probe ACCTGAATCACATCGAGCC ACTGA-BHQ1 Reverse 28 TTCTCGCTGAAGCTGAATT T FOXA2 Forward 29 CTA CTC CTC CGT GAG CAA 74 CAT GAA C Taqman 30 FAM-GCC TGG GGA TGA probe ACG GCA TGA ACA C-BHQ1 Reverse 31 GCC GCC GAC ATG CTC ATG TA MK167 Forward 32 TAA TGA GAG TGA GGG 87 AAT ACC TTT G Taqman 33 FAM-GGC GTG TGT CCT TTG probe GTG GGC A-BHQ1 Reverse 34 AGG CAA GTT TTC ATC AAA TAG TTC A NPTN Forward 35 ACC AGT GAA GAG GTC 88 ATT ATT CGA GAC A Taqman 36 FAM-CCT GTT CTC CCT GTC probe ACC CTG CAG TGT AAC- BHQ1 Reverse 37 TAT GTA AGG GTG TGA GAG CTG GAG GT sNA12 Forward 38 TGT GAC AAG GAA TAT 81 GTG AGC CTG G Taqman 39 FAM-CCT GAA GAT GCA probe TAT TCG GAC CCA CAC ATT-BHQ1 Reverse 40 CGC AGA TCT TGC AAA CAC AAG G TERT Forward 41 TGA CGT CCA GAC TCC GCT 83 TCAT Taqman 42 FAM-GCT GCG GCC GAT probe TGT GAA CAT GGA-BHQ1 Reverse 43 ACG TTC TGG CTC CCA CGA CGT A VIM Forward 44 ATG TTG ACA ATG CGT CTC 99 TGG CA Taqman 45 FAM-TGA CCT TGA ACG probe CAA AGT GGA ATC TTT GC- BHQ1 Reverse 46 ATT TCC TCT TCG TGG AGT TTC TTC AAA GAPDH Forward 47 CCA TCT TCC AGG AGC 90 GAG ATC C Taqman 48 FAM-TCC ACG ACG TAC probe TCA GCG CCA GCA-BHQ1 Reverse 49 ATG GTG GTG AAG ACG CCA GTG
Table 5 is a list of primer and probe sequences for all markers used in the present invention.

Comparative Example

[0093] Circulating tumor cells may exist in the blood in colorectal cancer or advanced adenoma, a precursor of colorectal cancer, and accordingly, an artificial intelligence algorithm-based model was constructed to determine the relative expression level of each group by targeting 10 genes (EpCAM, ERBB2, FOXA2, KRT19, MCAM, MKi67, NPTN, SNAI2, TERT, VIM)) known to have changes in relative expression level in circulating cancer cells, and to distinguish colorectal cancer or advanced adenoma from the normal group.

Collection of Clinical Specimens

[0094] From 2017 to 2022, Blood samples from subjects scheduled for colonoscopy were collected at the Shinchon Severance Hospital (Approval No. 4-2017-0148), the Gangnam Severance Hospital (Approval No. 3-2017-0024), the Kangbuk Samsung Hospital (Approval No. 2017-02-022-009) in the Department of Gastroenterology, and the Health Examination Center of Wonju Severance Christian Hospital (approval number CR319115) with the approval of the Bioethics Review Board (IRB) of each institution. A total of 3 ml of blood was collected using a Tempus blood tube (Applied Biosystems®). Subjects were classified as follows through the results of colonoscopy (Table 6)

TABLE-US-00006 TABLE 6 No. of samples Classification Classification criteria (persons) Colorectal As a result of colonoscopy, 148 cancer subjects with cancer in the colon group Advanced As a result of colonoscopy, 289 adenoma subjects with advanced adenoma group in the colon Normal Subjects with no lesions in the 142 group large intestine as a result of colonoscopy Total 579
Table 6 shows the classification of subjects and the number of samples according to colonoscopy results.
Isolation of Total RNA from Blood Specimens

[0095] Total RNA is isolated from a blood sample collected with a Tempus tube using the Tempus blood RNA isolation kit (Applied Biosystems®).

cDNA Construction from Isolated Total RNA and qPCR
i. Complementary DNA (cDNA) Synthesis
Isolated total RNA 1.5-4.5 ug, Random primer (3 ug/uL) (Invitrogen) 2.5 uL, dNTP mixture (2.5 mM each) (Intron) 2.5 uL, M-MLV reverse transcription polymerase (200 U/uL) (Invitrogen) 2.5 uL, 10 μL of 5× First-strand buffer (250 mM Tris-HCl) (Invitrogen), and 5 μL of Dithiothreitol (0.1 M) (Invitrogen) were added, and ultrapure water was added to a final volume of 50 μL, and mixed well. The synthetic reaction solution was reacted in a thermocycler (Applied Biosystems) at 25° C., 30 minutes—37° C., 50 minutes—70° C., 15 minutes to synthesize cDNA.
ii. Perform Quantitative Polymerase Chain Reaction (qPCR)
For the composition of the qPCR reaction, added 10 μL of THUNDERBIRD® Probe qPCR Mix (TOYOBO), Forward/Reverse Primer, Probe (10 pmole/uL) 1 μL, and added 2 μL of synthesized cDNA, and add ultrapure water to make the final volume 20 μL, and mixed. The qPCR reaction was performed using CFX96 (Biorad), and the reaction temperature conditions were as follows. After 95° C., 3 minutes, 95° C., 3 seconds—60° C., 30 seconds were repeated 40 times. Each time the annealing process (60° C., 30 seconds) was performed, a process of measuring fluorescence was added to measure the fluorescence value that increased by number of times. A constant fluorescence value was set as the threshold, and the Cq value, which is the number of cycles at the time of reaching the threshold, was derived.

Confirmation of Results and Analysis of Relative Expression of Target Genes

[0096] Using the Cq value of the GAPDH gene used as an endogenous control, the relative expression level (2−ΔCq) of the target gene is calculated using the Cq value of the target gene. A list of targeted genes follows (Table 7).

2.sup.−ΔCq=2.sup.−(target gene Cq−GAPDH gene Cq)  [Calculation formula]

TABLE-US-00007 TABLE 7 No. Blood genetic markers 1 EpCAM Epithelial Cell Adhesion Molecule 2 ERBB2 Erb-B2 Receptor Tyrosine Kinase 2 3 FOXA2 Forkhead Box A2 4 KRT19 Keratin 19 5 MCAM Melanoma Cell Adhesion Molecule 6 MKi67 Marker Of Proliferation Ki-67 7 NPTN Neuroplastin 8 SNAI2 Snail Family Transcriptional Repressor 2 9 TERT Telomerase Reverse Transcriptase 10 VIM Vimentin
Table 7 is a list of target blood genetic markers of comparative example.

Establishment of a Classification Model for the Purpose of Screening for Colorectal Cancer and Advanced Adenoma by Substituting the Relative Expression Level of Target Genes

[0097] An artificial intelligence algorithm-based classification model was constructed using the H2O package (version 3.32.1.3) of Statistical R software (version 3.6.3). The production of colorectal cancer and advanced adenoma diagnosis prediction models was based on Deep neural network (DNN), Generalized linear model (GLM), Random Forest (RF), and Gradient boosting machine (GBM) algorithms, and several types of models (GLM, RF, DNN, GBM, stacked ensemble (SE)) was performed by grafting Automated machine learning (AutoML) method to build a model suitable for data, but is not limited thereto.
By dividing the entire sample into a training set and a test set, and by substituting the results of the training set, an artificial intelligence algorithm-based classification model that can distinguish between the colorectal cancer group and the advanced glandular group compared to the normal group is constructed, and the performance of the built model is evaluated using the test set (FIG. 4).
When building a model using a training set, a 5-fold cross-validation technique is applied so that the training set is divided into 5 areas and so at the same time as learning the model, the performance of the model was verified using each area to build a high-performance model.
The performance of the artificial intelligence classification model was judged through the AUROC and AUPRC values of the training set and test set based on the AUROC and AUPRC values, which are representative performance indicators of the classification model. Among them, the model with the best performance was selected based on the performance of the new test set that was not used for model learning.
The AUROC and AUPRC values of the GLM, DNN, GBM, and RF models built based on each algorithm and the SE model built through AutoML are as follows (Table 8). As a result, the AUROC and AUPRC indicators were the highest in the RF and GBM model based on the test set.

TABLE-US-00008 TABLE 8 Training set Test set Model AUROC AUPRC AUROC AUPRC GLM 0.91 0.96 0.86 0.96 RF 0.90 0.96 0.94 0.98 DNN 0.99 1.00 0.92 0.97 GBM 1.00 1.00 0.94 0.98 AutoML 0.98 0.99 0.91 0.97 (SE)

[0098] Table 8 shows AUROC and AUPRC performance indicators in the training set and test set.

[0099] As a result of confirming the sensitivity and specificity of each group in the RF model and the GBM model, the sensitivity for distinguishing the colorectal cancer group in the RF model was 81.8% and the sensitivity for distinguishing the advanced adenoma group was 86.4% (Table 9). The specificity for classifying the normal group was 83.3%, the sensitivity for classifying the colorectal cancer group in the GBM model was 78.4%, the sensitivity for classifying the advanced adenoma group was 88.9%, and the specificity for classifying the normal group was 80.6% (Table 10). Therefore, an RF model with higher sensitivity for distinguishing colorectal cancer and higher specificity for distinguishing normal group was selected.

TABLE-US-00009 TABLE 9 Result of Test set (Total 154 persons) Positive Negative Sensitivity Specificity Classification (persons) (persons) (%) (%) Colorectal 100 18 84.7 cancer group + Advanced adenoma group(n = 118) Colorectal 30 7 81.1 cancer group (n = 37) Advanced 70 11 86.4 adenoma group(n = 81) Normal 6 30 83.3 group (n = 36)

[0100] Table 9 shows the sensitivity and specificity results for each group of the RF model.

TABLE-US-00010 TABLE 10 Result of Test set (Total 154 persons) Positive Negative Sensitivity Specificity Classification (persons) (persons) (%) (%) Color 101 17 85.6 ectal cancer group + Advanced adenoma group(n = 118) Colorectal 29 8 78.4 cancer group (n = 37) Advanced 72 9 88.9 adenoma group(n =81) Normal 7 29 80.6 group(n = 36)

[0101] Table 10 shows the sensitivity and specificity results for each group of the GBM model.