CN110577998A

CN110577998A - Construction of molecular model for predicting postoperative early recurrence risk of liver cancer and application evaluation thereof

Info

Publication number: CN110577998A
Application number: CN201910099925.9A
Authority: CN
Inventors: 夏强; 蔡杰; 孔晓妮; 邬海龙
Original assignee: Renji Hospital Shanghai Jiaotong University School of Medicine
Current assignee: Renji Hospital Shanghai Jiaotong University School of Medicine
Priority date: 2019-01-31
Filing date: 2019-01-31
Publication date: 2019-12-17

Abstract

The invention relates to construction of a molecular model for predicting early relapse risk after liver cancer operation and application evaluation thereof. Specifically, the invention discovers a biomarker set which comprises a plurality of biomarkers of early postoperative recurrence of liver cancer, can be used for evaluating the risk of early postoperative recurrence of a liver cancer patient, has the advantages of high sensitivity, high specificity and high accuracy, and has important application value. The invention also develops a risk prediction algorithm model and a device for early relapse after liver cancer operation for the first time, can assist clinical early prediction and early diagnosis of the early relapse after the liver cancer operation, implements more close postoperative follow-up and further auxiliary treatment, and greatly improves the postoperative prognosis of liver cancer patients.

Description

Construction of molecular model for predicting postoperative early recurrence risk of liver cancer and application evaluation thereof

Technical Field

The invention relates to the field of medical biological detection, in particular to construction and application evaluation of a molecular model for predicting postoperative early-stage recurrence risk of liver cancer.

Background

Hepatocellular carcinoma (HCC) is the most common primary liver cancer and the second leading cause of cancer-related death worldwide. Currently, the diagnosis and operation technology of liver cancer is gradually mature, but a large number of liver cancer patients frequently relapse after the operation and further progress to the advanced liver cancer, and finally the overall prognosis of liver cancer is poor. It is known that liver cancer recurrence is divided into early-stage recurrence and late-stage recurrence by taking 2 years after operation as a boundary, wherein the early-stage recurrence is considered as 'true recurrence' and accounts for more than 70% of liver cancer recurrence, and is a main reason for low postoperative survival rate of liver cancer patients. Therefore, if the liver cancer patients with high risk of early relapse can be identified as early as possible, more close postoperative follow-up and further adjuvant therapy are implemented, and the postoperative prognosis of the liver cancer patients can be greatly improved.

Currently, the TNM staging system is the most common method used to evaluate the prognosis after liver cancer surgery, but the prediction effect is not very ideal. Clinically, liver cancer patients with the same TNM stage have distinct postoperative prognosis, suggesting that additional staging factors are urgently needed to increase the predictive effect of the current staging system.

Therefore, there is an urgent need in the art to develop a reliable method for early prediction and diagnosis of the risk of postoperative recurrence of liver cancer.

Disclosure of Invention

The invention aims to provide a reliable method for early predicting and diagnosing the postoperative recurrence risk of liver cancer.

In a first aspect of the present invention, there is provided a use of a biomarker set and/or a detection reagent thereof for preparing a kit for diagnosis and/or prognosis evaluation of liver cancer, wherein the biomarker set comprises TMCO 3.

In another preferred embodiment, the prognostic assessment is a post-operative prognostic assessment.

in another preferred embodiment, the prognostic assessment comprises assessing the risk of liver cancer recurrence, preferably comprises assessing the risk of postoperative recurrence of liver cancer.

In another preferred example, the risk of liver cancer recurrence comprises risk of early stage recurrence of liver cancer and/or risk of late stage recurrence of liver cancer, preferably comprises risk of early stage recurrence after liver cancer surgery.

In another preferred example, the prognostic assessment includes predicting the likelihood that a liver cancer patient will not relapse with liver cancer.

In another preferred embodiment, the prognostic assessment comprises predicting the survival time of a liver cancer patient after receiving treatment, preferably predicting the survival time of a liver cancer patient after receiving surgical treatment.

In another preferred embodiment, any one member of the biomarker panel comprises a nucleic acid form (e.g. DNA, mRNA), and/or a protein form thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: GCGR, IGFALS, or a combination thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: GNE, RNF125, MCM4, or a combination thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: NSMCE2, AGL, CIDEB, DBH, or a combination thereof.

in another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: TARBP1, ID2, TOP2A, THY1, FAM189B, STAB2, LAMC1, DCXR, KIF20A, IGF1, AURKA, AURKB, RCAN1, ANXA10, or a combination thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: TARBP1, ID2, GCGR, or a combination thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: TOP2A, THY1, IGFALS, FAM189B, STAB2, or a combination thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: GNE, RNF125, LAMC1, MCM4, DCXR, KIF20A, or a combination thereof.

in another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: IGF1, AURKA, NSMCE2, AGL, AURKB, RCAN1, CIDEB, DBH, ANXA10, or a combination thereof.

In another preferred embodiment, the biomarker panel comprises biomarkers selected from the group consisting of: TMCO3, TOP2A, THY1, FAM189B, LAMC1, MCM4, KIF20A, AURKA, NSMCE2, AURKB, or combinations thereof.

In another preferred embodiment, the biomarker panel comprises biomarkers selected from the group consisting of: TARBP1, ID2, GCGR, IGFALS, STAB2, GNE, RNF125, DCXR, IGF1, AGL, RCAN1, CIDEB, DBH, ANXA10, or a combination thereof.

In another preferred embodiment, the biomarker panel comprises biomarkers selected from the group consisting of: TMCO3, TOP2A, THY1, FAM189B, LAMC1, MCM4, KIF20A, AURKA, NSMCE2, AURKB, TARBP1, ID2, GCGR, IGFALS, STAB2, GNE, RNF125, DCXR, IGF1, AGL, RCAN1, CIDEB, DBH, ANXA10, or a combination thereof.

in another preferred example, the set of biomarkers is TMCO 3.

in another preferred embodiment, the biomarker panel comprises two or more biomarkers selected from table a:

TABLE A

In another preferred embodiment, the set comprises biomarkers b1-b 24.

In another preferred embodiment, the set comprises the biomarkers bm and one or more biomarkers selected from the subset X consisting of the biomarkers b1-b (m-1) and b (m +1) -b24, m being an integer and 1. ltoreq. m.ltoreq.24, preferably 1. ltoreq. m.ltoreq.15, preferably 1. ltoreq. m.ltoreq.10, more preferably 1. ltoreq. m.ltoreq.5.

In another preferred embodiment, m is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24;

Wherein, when m is 1, the biomarker b1-b (m-1) is not present;

When m is 2, the biomarker b1-b (m-1) represents a biomarker b 1;

when m is an integer of 3 to 24, the subset X consists of biomarkers b1 to b (m-1) and biomarkers b (m +1) to b 24;

When m is 23, the biomarkers b (m +1) to b24 represent the biomarker b 24;

When m is 24, the biomarkers b (m +1) -b24 are not present.

In another preferred embodiment, the set comprises biomarker b4 and one or more biomarkers selected from subset Y, wherein subset Y consists of biomarkers b1-b3, b5-b 24.

In another preferred embodiment, said set comprises biomarkers b1, b2, b3, b4 and one or more biomarkers selected from subset Z, wherein subset Z consists of biomarkers b5-b 24.

In another preferred embodiment, said set comprises biomarkers selected from the group consisting of: b3, b4, b7, b10, b11, b13, b18, b19, b22, b23, or a combination thereof.

In another preferred embodiment, said set comprises biomarkers selected from the group consisting of: b1, b2, b3, b4, b5, b6, b7, b8, b9, or a combination thereof.

In another preferred embodiment, the biomarker or biomarker panel is derived from liver cancer tissue.

In another preferred example, the detection reagent is a reagent for detecting the expression level of the biomarker panel.

In another preferred embodiment, the kit further comprises an instruction, and the instruction describes a method for prognosis evaluation, wherein the method for prognosis evaluation comprises the following steps:

(1) Providing a sample from a patient suffering from liver cancer to be tested, and detecting the expression levels of the biomarkers in the sample, wherein the sample comprises TMCO 3;

(2) calculating a risk score for the biomarker based on its expression level; and

(3) comparing the risk score to a reference value to derive the prognostic evaluation result.

In another preferred example, the reference value includes normal tissue (e.g. liver tissue) from healthy population, or normal tissue, tissue beside cancer or liver cancer tissue from non-recurrent liver cancer patient population, and/or normal tissue or tissue beside cancer of the liver cancer patient to be tested.

in another preferred embodiment, the sample is selected from the group consisting of: liver cancer tissue.

In another preferred example, the risk score is calculated by formula I:

The risk score (- × TARBP expression state) + (× ID expression state) + (× GCGR expression state) + (× IGFALS expression state) + (× STAB expression state) + (× GNE expression state) + (× RNF125 expression state) + (× DCXR expression state) + (× IGF expression state + (× AGL expression state) + (× RCAN expression state) + (× CIDEB expression state) + (× DBH expression state) + (× anxaexpression state) + (× AURKB expression state) + (× mcte expression state) + (× AURKA expression state) + (× KIF20 expression state) + (MCM × expression state) + (× LAMC expression state) + (× FAM189 expression state) + (× THY expression state) + (× TOP2 expression state) + (× TMCO) (I).

wherein each biomarker expression state is 0 or 1, low expression state is equal to 0, and high expression state is equal to 1.

In another preferred embodiment, the step (2) further comprises the step of establishing a multivariate statistical model to output the risk of postoperative recurrence of liver cancer (preferably the risk of postoperative early recurrence), and preferably the multivariate statistical model is a LASSO regression model.

In another preferred embodiment, said liver cancer patient is judged to have a higher risk of postoperative recurrence of liver cancer if said risk score is higher.

in another preferred example, a significantly increased level of expression of one or more biomarkers selected from subset M, wherein subset M comprises biomarkers b4, b5, b6, b8, b12, b13, b15, b17, b18, and b20, when compared to a reference data set, indicates a liver cancer patient at risk of postoperative recurrence of liver cancer.

In another preferred example, the "significantly increased" means that the expression level of the biomarker C1 in the liver cancer patient is greater than or equal to 1.5, preferably greater than or equal to 2.0, and more preferably greater than or equal to 3, relative to the reference value C0, C1/C0.

In another preferred example, one or more biomarkers selected from subset N comprising biomarkers b1, b2, b3, b7, b9, b10, b11, b14, b16, b19, b21, b22, b23, and b24 are expressed at a significantly reduced level and a higher risk score when compared to a reference data set, indicating that the liver cancer patient is at risk of postoperative recurrence of liver cancer.

In another preferred example, the "significantly reduced" means that the ratio C0/C1 of the reference value C0 to the expression level C1 of the biomarker of the liver cancer patient is not less than 1.5, preferably not less than 2.0, more preferably not less than 3.

In another preferred embodiment, one or more biomarkers selected from subset H, including biomarkers b1-b24, are differentially expressed and a higher risk score indicates that the liver cancer patient is at risk of postoperative recurrence of liver cancer, as compared to a reference data set.

In another preferred example, the subset H comprises the biomarkers b1-b9, preferably the subset H comprises the biomarkers b1-b 4.

In another preferred embodiment, the fold difference of the differential expression is more than or equal to 1.5.

In another preferred embodiment, the expression level of the biomarker is detected by PCR detection, expression profiling chip detection or high throughput sequencing.

In another preferred embodiment, the PCR assay comprises QPCR, fluorogenic quantitative PCR, RT-PCR, or a combination thereof.

In a second aspect of the present invention there is provided a set of biomarkers comprising TMCO3 and one or more biomarkers selected from subset Y, wherein subset Y comprises: TARBP1, ID2, GCGR, TOP2A, THY1, IGFALS, FAM189B, STAB2, GNE, RNF125, LAMC1, MCM4, DCXR, KIF20A, IGF1, AURKA, NSMCE2, AGL, AURKB, RCAN1, CIDEB, DBH and ANXA 10.

In another preferred example, the plurality is 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 or 23.

In another preferred embodiment, the biomarker panel comprises two or more biomarkers selected from table a.

In another preferred embodiment, the set comprises biomarkers b1-b 24.

Wherein, when m is 1, the biomarker b1-b (m-1) is not present;

when m is 2, the biomarker b1-b (m-1) represents a biomarker b 1;

When m is 23, the biomarkers b (m +1) to b24 represent the biomarker b 24;

when m is 24, the biomarkers b (m +1) -b24 are not present.

In another preferred embodiment, said biomarker or biomarker panel is used for the diagnosis and/or prognostic assessment of liver cancer (patient).

In a third aspect of the invention, there is provided a set of biomarkers, said set comprising two or more biomarkers selected from the group consisting of: TMCO3, GCGR, IGFALS, or combinations thereof.

In another preferred embodiment, the biomarker panel further comprises a biomarker selected from the group consisting of: GNE, RNF125, MCM4, NSMCE2, AGL, CIDEB, DBH, or combinations thereof.

In a fourth aspect of the invention, there is provided a set of biomarkers, said set comprising two or more biomarkers selected from the group consisting of: TARBP1, ID2, GCGR, TMCO3, or a combination thereof.

In a fifth aspect of the invention, there is provided a combination of reagents for the prognostic assessment of a patient with liver cancer, the combination of reagents comprising reagents for detecting each biomarker in a collection as described in any one of the second to fourth aspects of the invention.

In another preferred embodiment, the reagents comprise reagents for detecting the expression level of each of the biomarkers.

In another preferred embodiment, the reagent comprises a reagent for detecting the expression level of each biomarker by PCR detection, expression profiling chip detection or high-throughput sequencing detection.

In a sixth aspect of the invention, there is provided a kit comprising a biomarker panel according to any one of the second to fourth aspects of the invention, and/or a combination of reagents according to the fifth aspect of the invention.

in another preferred embodiment, the kit further comprises an instruction describing a reference data set of the levels of the respective biomarkers in the set according to the second aspect of the invention derived from liver cancer tissue, paracancerous tissue and/or normal liver tissue of a population of patients with a non-relapsing liver cancer and/or of a healthy population.

In a seventh aspect of the present invention, there is provided a method for prognosis evaluation of a liver cancer patient, comprising the steps of:

(1) Providing a sample from a patient with liver cancer, and detecting the expression levels of the biomarkers in the sample from the collection, wherein the collection comprises TMCO 3;

In another preferred embodiment, the set further comprises one or more biomarkers selected from the group consisting of: TARBP1, ID2, GCGR, TOP2A, THY1, IGFALS, FAM189B, STAB2, GNE, RNF125, LAMC1, MCM4, DCXR, KIF20A, IGF1, AURKA, NSMCE2, AGL, AURKB, RCAN1, CIDEB, DBH, ANXA10, or a combination thereof.

In another preferred embodiment, the set further comprises biomarkers selected from the group consisting of: TARBP1, ID2, GCGR, or a combination thereof.

In another preferred example, the risk score is calculated by formula I:

In another preferred example, the judgment criterion of the expression state of the biomarker bm is as follows:

When the expression level Xm of the biomarker bm is more than or equal to the expression critical value Ym of the biomarker bm, the expression state of the biomarker bm is 1 (high expression state); and

When the expression level Xm of the biomarker bm < the expression cut-off Ym of the biomarker bm, the expression state of the biomarker bm is 0 (low expression state);

Wherein m is an integer and is more than or equal to 1 and less than or equal to 24.

in another preferred example, the method further comprises the steps of: the expression cutoff (Ym) for each biomarker bm was determined.

in another preferred example, the method further comprises the steps of: a ROC curve is plotted for the biomarkers and the expression cut-off (i.e. the optimal cut-off for low and high expression) for the biomarkers is determined based on the ROC curve.

In another preferred embodiment, the expression level of each biomarker is detected by PCR detection, expression profiling chip detection or high throughput sequencing.

In an eighth aspect of the present invention, there is provided a method for screening a candidate compound for reducing the risk of recurrence of liver cancer, comprising the steps of:

(1) Administering a test compound to a sample derived from a liver cancer patient in a test group, detecting the expression level V1 of each biomarker in the pool in the samples derived from said liver cancer patient in the test group; in a control group, a blank control (including vehicle) is administered to the sample derived from the liver cancer patient, and the level of each biomarker in the pool in the sample derived from the liver cancer patient in the control group is measured at V2; and

(2) comparing the level V1 and the level V2 detected in the previous step to determine whether the test compound is a candidate compound for reducing the risk of liver cancer recurrence, wherein the set comprises TMCO 3.

In another preferred embodiment, the candidate compound is a candidate compound that reduces the risk of early relapse after liver cancer surgery.

in another preferred example, a test compound is a candidate compound for the treatment of macular neovascular disease if the level V1 of one or more biomarkers selected from subset H is significantly lower than the level V2, wherein subset H comprises biomarkers b1-b 24.

In another preferred embodiment, the expression "significantly lower" means that the ratio of level V1/level V2 is ≦ 0.8, preferably ≦ 0.6, more preferably ≦ 0.4.

In a ninth aspect of the present invention, there is provided a method of establishing a model for evaluating the prognosis of a patient with liver cancer or predicting the risk of recurrence after a liver cancer surgery, the method comprising: identifying differentially expressed substances in the liver cancer tissue and the tissue beside the cancer,

Wherein the differentially expressed material comprises biomarkers in one or more biomarker panels, wherein the panel comprises TMCO 3.

in another preferred embodiment, the differentially expressed material comprises one or more biomarkers selected from subset H, wherein subset H comprises biomarkers b1-b 24.

In another preferred embodiment, the method is selected from the group consisting of: linear models, support vector machines, neural networks, classification regression trees, ensemble learning methods, linear discriminant analysis, neighbor methods, bayesian networks, independent component analysis, or combinations thereof.

The tenth aspect of the present invention provides an early-stage auxiliary screening system for liver cancer recurrence, which comprises:

(a) A biomarker expression level input module for inputting the expression level of each biomarker in a set in a sample derived from a subject, wherein the set comprises TMCO 3;

(b) The liver cancer recurrence judging and processing module is used for grading the expression level of the input biomarker according to a preset judgment standard so as to obtain a risk degree score; comparing the risk degree score with a risk degree threshold value of liver cancer recurrence to obtain an auxiliary screening result, wherein when the risk degree score is higher than the risk degree threshold value, the risk degree score indicates that the recurrence risk of the liver cancer of the target is higher than that of a non-recurrence liver cancer patient; and

(c) And the auxiliary screening result output module is used for outputting the auxiliary screening result.

in another preferred embodiment, the subject is a liver cancer patient.

In another preferred embodiment, the biomarker expression level input module comprises a sample collector and a gene expression level measuring device.

In another preferred embodiment, said score comprises (a) a score for the expression level of a single biomarker; and/or (b) the sum of scores for the expression levels of a plurality of biomarkers.

In another preferred example, in the processing module, the risk degree scoring processing is performed by formula I:

a risk score of (— x TARBP expression state) + (× ID expression state) + (× GCGR expression state) + (× IGFALS expression state) + (× STAB expression state) + (× GNE expression state) + (× RNF125 expression state) + (× DCXR expression state) + (× IGF expression state + (× AGL expression state) + (× RCAN expression state) + (× CIDEB expression state) + (× DBH expression state) + (× ANXA expression state) + (× AURKB expression state) + (× mce expression state) + (× AURKA expression state) + (× KIF20 expression state) + (MCM × expression state) + (× LAMC expression state) + (× FAM189 expression state + (× THY) + (× 2 expression state) + (× TMCO) (I);

In another preferred example, the method further comprises the steps of: determining an expression threshold value corresponding to each biomarker.

in another preferred embodiment, the liver cancer recurrence judging and processing module comprises a processor and a memory, wherein the memory stores threshold data or a model of the risk of liver cancer recurrence based on the expression level of the biomarker.

In another preferred example, the output module comprises a report system.

It is to be understood that within the scope of the present invention, the above-described features of the present invention and those specifically described below (e.g., in the examples) may be combined with each other to form new or preferred embodiments. Not to be reiterated herein, but to the extent of space.

drawings

FIG. 1 shows a screening of gene sets for universal differential expression in liver cancer. (A-L) volcanic plots of differentially expressed genes in the data set (Affymetrix, A-D; Agilent, E-H; Illumina, I-L). An X axis: log (log)₂(FC); y-axis: log (log)₁₀(FDR). Green: down-regulating a gene; gray: a non-differential gene; red: up-regulating the gene. Venn diagram cross-analysis of differentially expressed genes in each gene set in (M-O) Affymetrix, Agilent and Illumina sequencing platforms. Each ellipse corresponds to a data set, and the number of differentially expressed genes for each overlapping region is labeled at the corresponding location. If a gene is differentially expressed in 3 or more than 3 data sets at the same time, the gene is considered to be a differentially expressed gene in the sequencing platform. (Q) differential expressed gene Venn graph Cross-analysis of Affymetrix, Agilent and Illumina sequencing platforms.

Fig. 2 shows the construction of a LASSO risk model in GSE14520 for predicting early postoperative recurrence of liver cancer. The "glmnet" function returns a series of LASSO risk models. Each curve represents a variable and each vertical dashed line (not shown) represents a lambda value. And (3) an upper transverse coordinate: number of non-zero coefficient variables of lambda value, lower abscissa: sum of absolute values of non-zero coefficients of λ value, ordinate: a non-zero coefficient of the lambda value. (B) The "cv. The average cross validation error of the model is minimal when λ is 0.04698, i.e., log (λ) — 3.058. (C) At the chosen lambda value, a total of 24 genes with non-zero influence coefficients were obtained, TARBP1, ID2, GCGR, IGFALS, STAB2, GNE, RNF125, DCXR, IGF1, AGL, RCAN1, CIDEB, DBH, ANXA10, AURKB, NSMCE2, AURKA, KIF20A, MCM4, LAMC1, FAM189B, THY1, TOP2A and TMCO 3.

Fig. 3 shows the predictive value of evaluating the risk model in GSE 14520. (A) Waterfall plot of risk score height and postoperative early relapse. (B) Recurrence rate within 2 years after surgery in low and high risk groups. (C) Kaplan-Meier curve of early tumor-free survival rate in low and high risk groups. (D) The risk model was followed by ROC curves at 6, 12, 18 and 24 months post-surgery. (E) The risk model was compared to the AUC values of each single gene. (F) The risk model is compared to the AUC values of each clinical pathology index. (G) Multi-factor COX regression analysis of early tumor-free survival. P < 0.001.

Figure 4 shows the predictive value of validating the risk model in TCGA. (A) Waterfall plot of risk score height and postoperative early relapse. (B) Recurrence rate within 2 years after surgery in low and high risk groups. (C) Kaplan-Meier curve of early tumor-free survival rate in low and high risk groups. (D) The risk model was followed by ROC curves at 6, 12, 18 and 24 months post-surgery. (E) The risk model was compared to the AUC values of each single gene. (F) The risk model is compared to the AUC values of each clinical pathology index. (G) Multi-factor COX regression analysis of early tumor-free survival. P < 0.001.

Figure 5 shows an early relapse stratification analysis based on this risk model. (A) Early relapse stratification analysis in GSE14520 for this risk model. (B) Stratification of early relapse in TCGA for this risk model. (C-E) Kaplan-Meier curves for early tumor-free survival in a low, high risk group of a subgroup of TNMI, II and III stage liver cancer patients in GSE 14520. (F-H) Kaplan-Meier curves for early tumor-free survival in the low and high risk groups of the TNM I, II and III subgroups of liver cancer patients in TCGA.

Fig. 6 shows nomogram construction based on the risk model and its clinical application. (A) And integrating the risk model and the nomograms of the clinical and pathological indexes. (B) The nomograms of the clinical and pathological indexes are simply integrated. (C) Alignment curve of alignment chart a. (D) Alignment curve of alignment chart B. (E) Decision graphs for nomograms a and B.

Detailed Description

The inventor of the present invention has conducted extensive and intensive studies and unexpectedly found a novel biomarker for early recurrence after liver cancer surgery for the first time. Specifically, the invention discovers a biomarker set which comprises a plurality of biomarkers of early postoperative recurrence of liver cancer, can be used for evaluating the risk of early postoperative recurrence of a liver cancer patient, has the advantages of high sensitivity, high specificity and high accuracy, and has important application value. The invention also develops a risk prediction algorithm model and a device for early relapse after liver cancer operation for the first time, can assist clinical early prediction and early diagnosis of the early relapse after the liver cancer operation, implements more close postoperative follow-up and further auxiliary treatment, and greatly improves the postoperative prognosis of liver cancer patients. The present invention has been completed based on this finding.

Term(s) for

The terms used herein have meanings commonly understood by those of ordinary skill in the relevant art. However, for a better understanding of the present invention, some definitions and related terms are explained as follows:

As used herein, the term "comprising" or "includes" can be open, semi-closed, and closed. In other words, the term also includes "consisting essentially of …," or "consisting of ….

As used herein, the term "biomarker panel" refers to one biomarker, or a combination of two or more biomarkers.

As used herein, the terms "post-operative" and "post-operative" are used interchangeably.

According to the present invention, the expression level of the biomarker is detected by PCR detection, expression profiling chip detection or high throughput sequencing.

According to the invention, the reference data set refers to a training set.

According to the present invention, the training set and the validation set have the same meaning, as is known from the prior art. In one embodiment of the invention, the training set refers to a set of biomarker levels in a liver cancer patient and/or a healthy control biological sample, or to a set of biomarker levels in liver cancer tissue, paracancerous tissue, and/or normal tissue in a liver cancer patient. In one embodiment of the invention, a validation set refers to a data set used to test the performance of a training set. In one embodiment of the invention, the level of the biomarker may be represented as an absolute value or a relative value according to the method of determination. For example, when mass spectrometry is used to determine the level of a biomarker, the intensity of the peak may represent the level of the biomarker, which is the level of a relative value; when PCR is used to determine the level of a biomarker, the copy number of the gene or the copy number of a gene fragment may represent the level of the biomarker.

As used herein, the term "reference value" or "reference value" refers to a value that is statistically correlated with a particular result when compared to the analysis result. In a preferred embodiment, the reference value is determined from a statistical analysis of studies comparing the expression of biomarkers in normal or para-cancerous tissues to known clinical outcomes. Some of these studies are shown in the examples section herein. However, studies from the literature and user experience with the methods disclosed herein can also be used to produce or adjust the reference values. The reference value may also be determined by considering conditions and outcomes particularly relevant to the patient's medical history, genetics, age, and other factors.

In one embodiment of the invention, the reference value refers to a reference value or normal value of a healthy control. In another preferred example, the reference value includes normal tissue (e.g. liver tissue) from healthy population, or normal tissue, tissue beside cancer or liver cancer tissue from non-recurrent liver cancer patient population, and/or normal tissue or tissue beside cancer of the liver cancer patient to be tested. It is clear to those skilled in the art that in the case of a sufficiently large number of samples, a range of normal values (absolute values) for each biomarker can be obtained by means of testing and calculation methods. Therefore, when the levels of biomarkers are detected by methods other than PCR, the absolute values of the levels of these biomarkers can be directly compared with normal values, thereby evaluating the risk of postoperative (early) recurrence of liver cancer patients, and diagnosing or early diagnosing liver cancer. Statistical methods may also be used in the present invention.

As used herein, the term "biomarker," also referred to as a "biomarker," refers to a measurable indicator of a biological state of an individual. Such biomarkers can be any substance in an individual as long as they are related to a particular biological state (e.g., disease) of the subject, e.g., nucleic acid markers (e.g., DNA), protein markers, cytokine markers, chemokine markers, carbohydrate markers, antigen markers, antibody markers, species markers (species/genus markers) and functional markers (KO/OG markers), and the like. Biomarkers are measured and evaluated, often to examine normal biological processes, pathogenic processes, or therapeutic intervention pharmacological responses, and are useful in many scientific fields.

As used herein, the term "individual" refers to an animal, particularly a mammal, such as a primate, preferably a human.

As used herein, terms such as "a," "an," and "the" refer not only to the singular, but also include the general class that can be used to describe a particular embodiment.

As used herein, the term "expression" includes the production of mRNA from a gene or portion of a gene, and includes the production of protein encoded by an RNA or gene or portion of a gene, as well as the presence of a test substance associated with expression. For example, cDNA, binding of a binding partner (e.g., an antibody) to a gene or other oligonucleotide, protein or protein fragment, and chromogenic moieties of the binding partner are included within the scope of the term "expression". Thus, an increase in the density of half-spots on immunoblots such as western blots is also within the scope of the term "expression" based on biological molecules.

it should be noted that the explanation of the terms provided herein is only for the purpose of better understanding the present invention by those skilled in the art, and is not intended to limit the present invention.

Recurrence of liver cancer

currently, the TNM staging system is the most common method used to evaluate the prognosis after liver cancer surgery, but the prediction effect is not very ideal. Clinically, liver cancer patients with the same TNM stage have distinct postoperative prognosis, suggesting that additional staging factors are urgently needed to increase the predictive effect of the current staging system. A great deal of previous basic research has shown that in liver cancer, certain specific molecules are significantly related to the malignant phenotype of liver cancer, and can effectively predict early recurrence of liver cancer. However, due to their specificity and sensitivity, single markers often fail to accurately predict the post-operative prognosis of liver cancer patients. When multiple markers are used for comprehensive evaluation (including simple comprehensive evaluation and weighted comprehensive evaluation), the accuracy and specificity of evaluation can be further remarkably improved. Therefore, the integration of multiple molecular markers into the same model can significantly improve the prediction value of the postoperative prognosis of liver cancer.

A Least Absolute value convergence and Selection Operator (LASSO) is a regression analysis method that can simultaneously perform variable Selection and regularization to improve the prediction accuracy and interpretability of the statistical model generated by the regression analysis method. The algorithm is widely applied to a Cox proportional hazards regression model and is used for survival analysis of high-dimensional data. High-throughput Gene Expression databases (Gene Expression Omnibus, GEO) and Cancer and tumor Gene maps (The Cancer Genome Atlas, TCGA) are currently two major public databases, and researchers can freely obtain various Gene Expression data based on Expression profiling chip detection or high-throughput sequencing. Therefore, using these public mass data and novel bioinformatic analysis methods, it is feasible and of great clinical significance to identify and construct more powerful and practical molecular marker models to predict early recurrence of liver cancer.

In the invention, all liver cancer gene expression data sets in the GEO database are searched, and 12 data sets meeting the standard are screened. After cross-analysis of differentially expressed genes in these 12 GEO datasets, a set of genes that are differentially expressed in liver cancer at large was identified. And (3) further carrying out LASSO regression analysis on the candidate differential expression genes by using the prognosis data of GSE14520 to construct a risk model consisting of 24 genes so as to predict the early relapse of the liver cancer. In addition to GSE14520, this risk model is also effective in predicting early postoperative recurrence of liver cancer in TCGA. Further layering analysis shows that the risk model can enhance the prediction value of TNM staging and BCLC staging on early relapse after liver cancer operation. Finally, for better clinical application, a nomogram based on the risk model is established, and the result shows that the nomogram can more accurately and effectively predict the early relapse of the liver cancer after the integration of the risk model.

Biomarkers of the invention

ROC-AUC

a method for evaluating model accuracy, wherein an ROC curve is a coordinate graph formed by a Receiver operating characteristic curve (Receiver operating characteristic curve) and a False positive probability (False positive rate) as a horizontal axis and a True positive probability (True positive rate) as a vertical axis, and the ROC curve is a comprehensive index reflecting sensitivity and specificity continuous variables. AUC is the Area under the ROC curve (Area under the curve). The ROC-AUC value is between 1.0 and 0.5, the closer to 1, the better the diagnosis effect is, the lower the accuracy is at 0.5-0.7, the certain accuracy is at 0.7-0.9, and the higher the accuracy is at AUC above 0.9. When AUC is 0.5, the diagnostic method is completely ineffective and is not valuable. AUC <0.5 does not correspond to the real case and occurs rarely in practice.

evaluation method

In another preferred embodiment, the method of the present invention can calculate the risk score by the formula:

Risk score S ═ W1S1+ W2S2+ WiS3+ … … WnSn;

Wherein W1 and W2 … … Wn are coefficients;

s1, S2 … … Sn are scores for each biomarker.

preferably, the coefficients and scores are as shown in table B.

TABLE B

In another preferred example, in the processing module, the risk degree scoring processing is performed as follows:

A risk score of (— x TARBP expression state) + (× ID expression state) + (× GCGR expression state) + (× IGFALS expression state) + (× STAB expression state) + (× GNE expression state) + (× RNF125 expression state) + (× DCXR expression state) + (× IGF expression state + (× AGL expression state) + (× RCAN expression state) + (× CIDEB expression state) + (× DBH expression state) + (× ANXA expression state) + (× AURKB expression state) + (× mcte expression state) + (× AURKA expression state) + (× KIF20 expression state) + (MCM × expression state) + (× LAMC expression state) + (× FAM189 expression state) + (× THY 2 expression state) + (× TMCO);

When a biomarker in the set of biomarkers employed is absent, the corresponding risk score sub-term may be deleted from the risk score formula.

For example, the judgment criteria of the expression state of the biomarker b1 are as follows:

When the expression level X1 of the biomarker b1 is more than or equal to the expression critical value Y1 of the biomarker b1, the expression state of the biomarker b1 is 1 (high expression state); when the expression level of the biomarker b1, X1 < the expression threshold Y1 of the biomarker b1, then the biomarker b1 expression status is 0 (low expression status).

In the method of the present invention, the expression threshold value corresponding to each biomarker bm may be determined to be Ym, m is an integer and m is 1. ltoreq. m.ltoreq.24 using the ROC curve.

Specifically, a ROC curve is plotted for a biomarker, and when the sum of sensitivity and specificity is maximized, the optimal cut-off for low and high expression of the biomarker is determined. When the expression level of the marker is higher than this cut-off value, the corresponding score of the expression status of the marker is 1 (high expression status); otherwise, it is marked as 0 (low expression state).

In a preferred embodiment, when the risk score of the subject is greater than or equal to the risk threshold of early recurrence after prediction surgery S0, it indicates that the subject has a high risk of liver cancer (early) recurrence.

in the methods of the invention, the ROC curve can be used to determine a cutoff value S0 for predicting the risk of early recurrence after surgery.

The experimental result of the invention shows that the marker of the invention can more accurately and effectively predict the early relapse of the liver cancer after the operation.

The main advantages of the invention include:

(a) the invention discovers a biomarker set for the first time, takes genes as biomarkers for early screening, prejudgment and prognosis evaluation of liver cancer, has the advantages of high sensitivity, high specificity and high accuracy, and has important application value.

(b) The invention discovers for the first time that TMCO3 can be used as a biomarker for diagnosis and prognosis evaluation of liver cancer.

(c) the invention also develops a risk prediction algorithm model and a device for early relapse after liver cancer operation for the first time, can assist clinical early prediction and early diagnosis of the early relapse after the liver cancer operation, implements more close postoperative follow-up and further auxiliary treatment, and greatly improves the postoperative prognosis of liver cancer patients.

The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Experimental procedures without specific conditions noted in the following examples, generally followed by conventional conditions, such as Sambrook et al, molecular cloning: the conditions described in the laboratory Manual (New York: Cold Spring harbor laboratory Press,1989), or according to the manufacturer's recommendations. Unless otherwise indicated, percentages and parts are percentages and parts by weight.

Materials and methods:

1. Preparation of liver cancer gene expression data set

Gene expression datasets for liver cancer were downloaded from GEO (https:// www.ncbi.nlm.nih.gov/GEO /) and TCGA (http:// cancer. nih. gov /) databases. In the GEO database, a total of 837 datasets were searched, of which 625 were human expression datasets. The appropriate GEO dataset was further screened using strict exclusion criteria: 1. absence of non-tumor samples; 2. the number of differentially expressed genes is less than 100; 3. the annotated genes were too few to account for 90% of the total transcriptome gene. The data sets that meet the criteria are then sorted by number of samples. Affymetrix, Agilent and Illumina are the most commonly used 3 sequencing platforms at present, and the data of different sequencing platforms are often different. In order to screen out genes which are generally differentially expressed in liver cancer, 4 data sets with the largest sample size are respectively selected from the three sequencing platforms, 12 data sets are obtained in total, and the differentially expressed genes are subjected to cross analysis. Details of these 12 data sets are shown in table 1. Among them, GSE14520 is the only expression data set with complete prognostic data. In addition, 215 liver cancer patients in the TCGA database had complete gene expression and postoperative recurrence data. The online analysis tool GEO2R was used to screen out differentially expressed genes (p-value <0.05 and fold difference >1.5) between liver cancer and paracancerous tissues in each dataset. Venn mapping and cross analysis were performed using the R language package "VennDiagram".

TABLE 1

Establishment of LASSO regression model

Of the 12 data sets, only GSE14520 had complete prognostic data. Therefore, a LASSO regression model was constructed using GSE 14520. Firstly, carrying out single-factor Cox regression analysis on each candidate differential expression gene and calculating a p value, drawing an ROC curve of each candidate differential expression gene and calculating an AUC value of the ROC curve, and when the p value is less than 0.25 and the AUC is more than or equal to 0.55, considering that the gene is related to early recurrence of the liver cancer, and carrying out further analysis to construct a LASSO regression model. The R language package "glmnet" returns a series of lambdas (λ s) values and models. And further performing ten-fold cross validation by using a 'cv. glmnet' function in the 'glmnet' packet, and selecting a lambda (lambda s) value with the minimum average cross validation error, wherein the corresponding LASSO model is the constructed ideal LASSO risk model. The risk score of early postoperative relapse of each liver cancer patient can be calculated through a risk model formula.

3. statistical analysis

And drawing an ROC curve according to the risk score of each patient, wherein the risk score corresponding to the maximum sum of sensitivity and specificity is a critical value for classifying the liver cancer patients into groups with low and high postoperative early relapse. The area under the ROC curve, i.e., the AUC value, is used to evaluate the prediction accuracy of the different variables. Survival curves were drawn using the Kaplan-Meier method and the log rank test was used to compare the survival differences between the different groups. The Pearson chi-square test or Fisher's exact test was used to examine the correlation between the risk model and each clinical pathology feature. Single and multifactorial Cox analyses were performed in sequence to investigate whether this risk model was independent of other clinical pathology features. A p value <0.05 was considered statistically significant.

4. Construction of alignment graph

Factors with p-value <0.05 (including risk models) were included to construct nomograms predicting early postoperative recurrence of liver cancer based on single-factor Cox prognostic analysis. A calibration curve and a consistency index (C-index) were used to evaluate the validity of the nomogram. Decision curve analysis was used to assess the clinical benefit of the nomograms. The analysis was done using the R language package "rms".

example 1 screening of Gene sets universally differentially expressed in liver cancer

The GEO database (http:// www.ncbi.nlm.nih.gov/GEO /) is a common gene expression database, currently having over 94000 datasets and over 2 million samples. These data sets have been widely used for biomarker identification and molecular typing, etc. of diseases by appropriate analytical methods and analytical tools. However, the samples or assay platforms incorporated into each dataset are different, and the results of the analysis vary widely from dataset to dataset. Therefore, cross-analyzing the results of multiple data sets will avoid such selective bias and make the results more objective.

In this example 12 liver cancer datasets from 3 most commonly used sequencing platforms (Affymetrix, Agilent and Illumina) were analyzed, each sequencing platform containing 4 datasets (table 1). Differentially expressed genes were identified separately for each dataset using GEO2R and are shown in the volcano plots in fig. 1A-1L. In addition, each sequencing platform comprises data sets and differential expression genes are subjected to cross analysis, and if a certain gene is differentially expressed in 3 or more data sets at the same time, the gene is considered to be the differential expression gene in the sequencing platform. Venn diagram shows that 1773, 726 and 870 differentially expressed genes were identified in the Affymetrix, Agilent and Illumina platforms, respectively (FIG. 1M-1O). Finally, the differentially expressed genes of the three sequencing platforms were further cross-analyzed, and 257 genes were screened out to be significantly differentially expressed in the three sequencing platforms, i.e., the genes were considered to be differentially expressed in liver cancer (fig. 1P).

example 2 construction of a LASSO Risk model for prediction of early postoperative recurrence of liver cancer in GSE14520

Of the 12 GEO data sets included, GSE14520 was the only one with complete clinical prognosis data. Therefore, based on the 257 candidate differentially expressed genes, the present example seeks to construct a risk model capable of effectively predicting early relapse after liver cancer surgery in GSE 14520. First, a ROC curve was plotted according to the expression value of each gene, and the critical expression value and AUC value of each gene were determined. The 221 patients in GSE14520 were classified into low and high expression states of each gene according to the cutoff value of the gene. Next, Cox single factor analysis was performed for each gene, and a p-value was calculated. When a certain gene simultaneously satisfies AUC ≥ 0.55 and p value <0.25, the gene is included in the next LASSO risk model construction. Finally, 112 genes meet the above conditions, 30 genes are up-regulated and 82 genes are down-regulated in liver cancer.

the R language package "glmnet" returns a series of models (fig. 2A), each different lambda value representing a LASSO risk model. Further using the ten-fold cross-validation method to select the optimal model, the results show that the average cross-validation error for this model is minimal when λ is 0.04698, i.e., log (λ) — 3.058 (fig. 2B). At this lambda value, a total of 24 genes with non-zero influence coefficients were obtained, TARBP1, ID2, GCGR, IGFALS, STAB2, GNE, RNF125, DCXR, IGF1, AGL, RCAN1, CIDEB, DBH, ANXA10, AURKB, NSMCE2, AUA, KIF20A, MCM4, LAMC1, FAM189B, THY1, TOP2A and TMCO3 (FIG. 2C). The risk model formula obtained is as follows: the risk score (- × TARBP expression state) + (× ID expression state) + (× GCGR expression state) + (× IGFALS expression state) + (× STAB expression state) + (× GNE expression state) + (× RNF125 expression state) + (× DCXR expression state) + (× IGF expression state + (× AGL expression state) + (× RCAN expression state) + (× cidb expression state) + (× DBH expression state) + (× ANXA expression state) + (× AURKB expression state) + (× mcte expression state) + (× AURKA expression state) + (× KIF20 expression state) + (MCM × expression state) + (× LAMC expression state) + (× FAM189 expression state) + (× THY 2 expression state) + (× TMCO) In (3), the low expression state of the gene is equal to 0 and the high expression state is equal to 1.

example 3 evaluation of the predictive value of the Risk model in GSE14520

The risk score for each patient in GSE14520 was calculated according to a risk model formula, as shown in fig. 3A, liver cancer patients are more prone to early relapse after surgery as the risk score increases. The optimal cut-off value was determined by plotting a ROC curve as-0.2600, based on which patients were classified into low and high risk groups. As shown in fig. 3B, the early postoperative recurrence rate of the low and high risk groups showed a significant difference at 6 months after surgery (1.65% in the low risk group and 25% in the high risk group, with a p value <0.001), which continued to exist and gradually increased with time. The Kaplan-Meier curve shows significantly poorer early tumor-free survival in patients in the high risk group compared to the low risk group (risk ratio 7.954, 95% confidence interval: 4.596-13.767, p-value 2.162E-18) (fig. 3C). Furthermore, the ROC curve shows that the risk model has excellent predictive value at 6 months, 12 months, 18 months and 24 months after surgery, with AUC values of 0.837, 0.843, 0.835 and 0.846(p values all <0.001) (fig. 3D). Further, the predictive value of the risk model was significantly greater compared to any single gene in the risk model or any other clinical pathology index (fig. 3E-3F).

Furthermore, the correlation analysis results showed that the risk model was positively correlated with multiple progressive clinical pathology features, including cirrhosis (p-value ═ 0.041), high AFP (p-value <0.001), late TNM stage (p-value <0.001), late BCLC stage (p-value <0.001), and late CLIP stage (p-value <0.001) (table 2). Multifactorial Cox regression analysis showed that the risk model was still significantly associated with early relapse after balance with gender, cirrhosis, TNM staging, BCLC staging and CLIP staging (risk ratio 6.811, 95% confidence interval: 3.799-12.214, p value <0.001, fig. 3G and table 3), suggesting that the risk model is an independent risk factor for early postoperative tumor-free survival.

TABLE 2

TABLE 3

Example 4 validation of predictive value of the risk model in TCGA

Further, the TCGA tests whether the risk model has similar prediction value in different liver cancer populations. Similarly, the risk score for each patient in TCGA was first calculated according to the risk model formula, and similar to GSE14520, patients with higher risk scores were more likely to develop early recurrence of liver cancer (fig. 4A). The optimal cutoff for this cohort was determined by the ROC curve to be-1.0522 and patients were assigned to low and high risk groups. As shown in fig. 4B, the relapse rates were 3.01%, 6.77%, 9.77% and 13.53% at 6 months, 12 months, 18 months and 24 months in the low-risk group patients, and 20.73%, 37.80%, 47.56% and 50% at 6 months, 12 months, 18 months and 24 months in the high-risk group patients, respectively, all significantly higher than that in the low-risk group (p-value < 0.001). In addition, high risk score patients had significantly poorer early postoperative tumor-free survival than low risk score patients (risk ratio: 5.982, 95% confidence interval: 3.414-10.480, p-value: 1.734E-12) (fig. 4C). The ROC curves show that the AUC values of the risk model in TCGA at 6 months, 12 months, 18 months and 24 months are 0.746, 0.773, 0.755 and 0.759, respectively (p values are all <0.001, fig. 4D). Likewise, the predictive value of this risk model was higher than any single gene or clinical pathology (p values <0.05) (fig. 4E-4F).

Correlation analysis showed that the risk model had a positive correlation with multiple progressive clinical pathology features including vascular invasion (p-value 0.023), high AFP (p-value 0.002) and R1 surgical margin (p-value 0.015) (table 4). Multifactorial Cox regression analysis showed that this risk model is also an independent risk factor for early postoperative tumor-free survival in TCGA (risk ratio 6.268, 95% confidence interval: 3.246-12.104, p value <0.001, fig. 4G and table 5).

TABLE 4

TABLE 5

example 5 early relapse stratification analysis based on this risk model

In order to further determine the range of the risk model suitable for the liver cancer population, the embodiment performs the hierarchical analysis of the early relapse. First, GSE14520 patients were stratified by each clinical pathology (sex, age, ALT, tumor size, foci, cirrhosis, AFP, TNM staging, BCLC staging and CLIP staging), and the prognostic analysis showed that this risk model was effective in predicting early postoperative recurrence of liver cancer in almost all subgroups (fig. 5A). Similarly, similar results were found in the TCGA cohort (fig. 5B), indicating that this risk model is applicable to the prediction of early postoperative recurrence in almost all patients with liver cancer.

Currently, most staging systems based on clinical pathological features, such as TNM staging and BCLC staging, are often used to differentiate postoperative prognosis in patients with liver cancer. However, patients who are clinically common in the same stage show different postoperative prognosis results, which indicates that the current staging system based on clinical pathological characteristics has certain limitations and needs to be perfected. Based on this, the risk model is further applied to patients in the same TNM stage for early recurrence prediction, and the results show that the risk model can effectively distinguish postoperative early recurrence of the patients in the same TNM stage, which indicates that the risk model can significantly enhance the prediction value of TNM stage on postoperative early recurrence of liver cancer (fig. 5C-5H).

Example 6 nomogram construction based on the risk model and clinical applications thereof

To facilitate clinical application, a nomogram was constructed based on the risk model and integrating various clinical pathological features (gender, cirrhosis, TNM stage, BCLC stage, and CLIP stage) (fig. 6A). Similarly, a nomogram that simply integrates clinical and pathological indices was constructed at the same time (fig. 6B). The calibration curve shows that the nomograms based on the risk model have a significantly higher degree of fit between predicted and actual outcomes (C-index 0.883, 95% confidence interval 0.836-0.929, fig. 6C) (p-value <0.001) compared to the nomograms of the simple clinical pathology indices (C-index 0.762, 95% confidence interval: 0.698-0.826, fig. 6D). In addition, decision curve analysis results show that risk model-based nomograms exhibit higher overall clinical net benefit than simple clinicopathologic index nomograms (fig. 6E).

Discussion:

postoperative recurrence of tumors remains a major obstacle to increasing postoperative prognosis in patients with tumors. At present, postoperative recurrence of liver cancer is divided into early and late recurrence by taking 2 years as a boundary. Among them, late-stage recurrence is considered to be a new tumor of the residual diseased liver, independent of the malignant nature of the original resected tumor. In contrast, early recurrence is often caused by intrinsic stealth intrahepatic metastases, and is closely related to the malignant properties of the original resected tumor, such as tumor size, foci, microvascular invasion and high AFP. Of the two recurrence types, early recurrence accounts for over 70% of the recurrence after liver cancer operation, and is the main reason for low postoperative survival rate of liver cancer patients. Therefore, if patients who are likely to suffer from postoperative early relapse can be identified as early as possible, and close follow-up and early intervention are performed, the postoperative prognosis of liver cancer patients can be greatly improved.

Currently, the TNM staging system and other prognosis scoring systems such as BCLC score, CLIP score and JIS score have been widely used to evaluate the postoperative prognosis of liver cancer patients. However, since these scoring systems only focus on clinical pathological features and ignore differences in the molecular genetic background of liver cancer, they often cannot be reliably and effectively predicted for the postoperative prognosis of liver cancer patients. Previous studies have shown that part of the genes can affect the malignant phenotype of liver cancer and can predict early postoperative recurrence of liver cancer. However, due to the diversity and systematicness of the tumor molecular genetic background, a single molecular marker often cannot accurately predict the early postoperative recurrence of liver cancer. Therefore, a prediction model of the multi-molecular marker is established by applying a specific analysis method to perfect the current tumor staging system, so that the postoperative early-stage recurrence high-risk patient can be identified as early as possible, and closer postoperative follow-up and early intervention can be performed to improve the postoperative prognosis of the liver cancer patient.

in the research, a plurality of statistical analysis methods are applied to analyze the public liver cancer gene expression data on the network, and a risk model consisting of a plurality of molecular markers is established to predict the early postoperative recurrence of liver cancer. First, a gene set with generally differential expression in liver cancer was identified by using GEO2R on-line analysis tool and Venn diagram cross-analysis, and comprised 257 genes. Secondly, carrying out single-factor Cox regression analysis and ROC curve drawing on each gene to calculate AUC values, and screening out 112 early relapse related genes. And finally, constructing a series of risk models by using the R language package 'glmnet', selecting the optimal risk model with the minimum average cross validation error after ten-fold cross validation, wherein the risk model formula consists of 24 genes and corresponding non-zero coefficients. The risk model is proved to be capable of effectively predicting the early postoperative recurrence of the liver cancer in the GSE14520 and TCGA queues, and the prediction value of the risk model is higher than that of any single gene or clinical and pathological index. The stratified analysis shows that the risk model can definitely increase the prediction value of early postoperative recurrence of the liver cancer in the TNM stage. In order to better realize clinical application, the nomogram is further constructed based on the risk model, a calibration curve shows that the prediction accuracy of the nomogram integrated with the risk model is higher, and decision curve analysis proves that the nomogram integrated with the risk model can obviously improve the overall clinical profitability of liver cancer patients.

In a word, the invention constructs a practical risk model which can effectively predict the early relapse of the liver cancer, and is a powerful supplement to the current TNM staging. Doctors or patients can individually predict the early recurrence risk of liver cancer through the risk model formula and make the next diagnosis and treatment plan, so that the social and national medical resources are more properly utilized and the current great trend of personalized medicine is met.

all documents referred to herein are incorporated by reference into this application as if each were individually incorporated by reference. Furthermore, it should be understood that various changes and modifications of the present invention can be made by those skilled in the art after reading the above teachings of the present invention, and these equivalents also fall within the scope of the present invention as defined by the appended claims.

Claims

1. Use of a biomarker panel and/or detection reagents thereof for the preparation of a kit for the diagnosis and/or prognostic evaluation of liver cancer, wherein the biomarker panel comprises TMCO 3.

2. The use of claim 1, wherein the biomarker panel further comprises a biomarker selected from the group consisting of: GCGR, IGFALS, or a combination thereof. .

3. A biomarker set comprising TMCO3 and one or more biomarkers selected from subset Y, wherein subset Y comprises: TARBP1, ID2, GCGR, TOP2A, THY1, IGFALS, FAM189B, STAB2, GNE, RNF125, LAMC1, MCM4, DCXR, KIF20A, IGF1, AURKA, NSMCE2, AGL, AURKB, RCAN1, CIDEB, DBH and ANXA 10.

4. A set of biomarkers, wherein said set comprises two or more biomarkers selected from the group consisting of: TMCO3, GCGR, IGFALS, or combinations thereof.

5. A set of biomarkers, wherein said set comprises two or more biomarkers selected from the group consisting of: TARBP1, ID2, GCGR, TMCO3, or a combination thereof.

6. A combination of reagents for the prognostic assessment of liver cancer patients, said combination of reagents comprising reagents for the detection of each biomarker in the collection according to any one of claims 3 to 5.

7. A kit comprising a biomarker panel according to any one of claims 3 to 5, and/or a combination of reagents according to claim 6.

8. A method for the prognostic evaluation of a patient with liver cancer, comprising the steps of:

9. A method of screening for a candidate compound that reduces the risk of recurrence of liver cancer, comprising the steps of:

10. An early-stage auxiliary screening system for recurrence of liver cancer, comprising: