Prediction method for progression-free lifetime of middle and advanced HCC patient after triple immunotherapy

Abstract

The invention discloses a method for predicting progression-free lifetime of middle and advanced HCC patients after triple immunotherapy. The method comprises the following steps: constructing a personalized tumor variation map for sequencing data of tumor tissues and paired white blood cells of the middle and advanced HCC patients; calculating a gene risk score by using a first model based on tumor mutation information in the personalized tumor mutation map; calculating a total risk score by using a second model by taking the gene risk score and the clinical variable as independent variables; and carrying out risk stratification according to the cutoff value of the total risk score, and predicting the probability that the patient accepts the PFS at different times of triple immunotherapy. The method can effectively predict progression-free survival conditions of middle and advanced hepatocellular carcinoma patients after triple immunotherapy, so that potential beneficial patients are screened out.

Description

Technical Field

[0001] The present application belongs to the field of biological detection information analysis, and specifically relates to a method for predicting progression-free survival of patients with advanced HCC after triple immunotherapy. Background Art

[0002] Liver cancer is still a huge challenge worldwide. Hepatocellular carcinoma (HCC) is the most common subtype of liver cancer, accounting for about 90% of the total number of cases. Most liver cancer patients have lost the opportunity for surgical treatment when diagnosed, and non-surgical local treatment and systemic treatment are the main treatments. Therefore, conversion therapy can achieve downstaging to resectable liver cancer stages for patients in the middle and late stages who are not surgically treatable. In the IMbrave150 study, atezolizumab (anti-PD-L1 inhibitor) and bevacizumab (anti-VEGFA; anti-angiogenic agent) ("T+A" therapy) had more significant survival benefits compared with traditional sorafenib in the first-line treatment of patients with middle and late-stage HCC. In addition, anti-vascular drugs combined with immunotherapy and interventional local chemotherapy (triplet therapy) are developing rapidly, which can further significantly improve progression-free survival (PFS) compared with "T+A" therapy, thus becoming the main method of conversion therapy for potentially resectable liver cancer.

[0003] However, not all patients can benefit from triple therapy. Clinical cohort data show that about 25% of advanced HCC patients have a progression-free survival (PFS) of less than three months. Therefore, patients with advanced HCC are a heterogeneous group with different subgroups, and there are differences in the efficacy and prognosis of triple therapy, and this heterogeneity may be attributed to differences in the endogenous characteristics of the tumor. Therefore, the current clinical difficulty lies in how to predict and screen patients who can benefit from this therapy before medication. In the existing technology, traditional clinical factors have no significant effect in predicting the efficacy and longer PFS of patients with advanced HCC. In addition, although there are a lot of studies and evidence in the current clinical technology or biomarkers (such as PD-1 / PD-L1, etc.) for predicting the efficacy of immunotherapy, there is a lack of relevant clinical evidence in predicting the efficacy of triple therapy for advanced HCC and screening the population that benefits. On the other hand, the prediction of the efficacy of immunotherapy, such as the detection of PD-1 / PD-L1, often requires tissue biopsy, which has limitations such as heterogeneity and invasiveness of tumor tissue. Therefore, it also leads to the barrenness of existing technology for predicting the efficacy of HCC triple therapy.

[0004] Exploratory studies on predictive biomarkers for triple therapy are also underway, and a recent study has highlighted potential predictive biomarkers in peripheral blood, such as CD4+ and CD8+ T cells. Clinical data from the present invention show that high infiltration rates of T cells and DC cells in tumors in baseline biopsies are associated with better prognosis. However, these studies have limitations such as low sample size and single center. Therefore, few effective biomarkers can be used in clinical practice. Current studies have shown that the cumulative effect of gene mutations in key biological signaling pathways may be used to predict the efficacy of immune checkpoint inhibitors (ICI). The R language package "PMAPscore" evaluates the level of signaling pathway mutation perturbation by combining the cumulative effect and position effect of gene mutations, which can more comprehensively and effectively evaluate the gene mutation effect of signaling pathways. The algorithm has been applied to clinical cohorts of non-small cell lung cancer (NSCLC) and melanoma treated with ICI, and correspondingly identified relevant signaling pathways to construct a prognostic model for predicting overall survival (OS), but OS is easily affected by subsequent treatment methods, and there are limitations in the efficacy and prognosis prediction after triple therapy. There is currently no effective biomarker or computational model to predict the prognosis of patients with advanced HCC receiving triple therapy. In addition, compared with predicting OS, PFS (progression-free survival) can more accurately reflect the clinical benefits of triple therapy in patients with advanced HCC. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting progression-free survival of patients with advanced HCC after triple immunotherapy, thereby screening out patients who may potentially benefit.

[0006] To achieve the above object, the present invention adopts the following technical solution:

[0007] Methods for predicting progression-free survival in patients with advanced HCC after triple immunotherapy include:

[0008] Construct a personalized tumor mutation map based on the sequencing data of tumor tissues and paired leukocytes from patients with advanced HCC;

[0009] Based on the tumor mutation information in the personalized tumor variation map, the first model is used to calculate the genetic risk score. Then, the second model is used to calculate the total risk score with the genetic risk score and clinical variables as independent variables. Risk stratification is performed according to the cutoff value of the total risk score to predict the probability of PFS of patients receiving triple immunotherapy at different times.

[0010] Among them, the first model and the second model are constructed with a retrospective cohort of patients with advanced HCC as training data, including: constructing a personalized tumor variation map based on tumor tissues of patients in the retrospective cohort before triple immunotherapy, calculating the perturbation level of tumor mutation information in the signal pathway by combining the cumulative effect and position effect of mutant genes, screening gene mutation perturbation pathways significantly associated with PFS, using the gene mutation perturbation pathways significantly associated with PFS as biomarkers, and constructing the first model based on regression analysis;

[0011] The second model was established based on regression analysis with the calculation results of the first model and the clinical variables of the retrospective cohort patients as independent variables, wherein the clinical variables were clinical risk factors significantly associated with PFS.

[0012] As a preferred embodiment, univariate Cox regression analysis and multivariate Cox stepwise regression analysis are used to screen gene mutation perturbation pathways significantly associated with PFS and establish the first model.

[0013] As a preferred implementation, the calculation results of the first model and the clinical variables of the retrospective cohort patients are used as independent variables, and a second model is established based on a nomogram.

[0014] As a preferred implementation, clinical risk factors significantly associated with PFS were screened out through univariate Cox regression analysis and multivariate Cox stepwise regression analysis.

[0015] Furthermore, the clinical risk factors include aspartate aminotransferase level AST, Child-Pugh grade and metastatic status.

[0016] As a preferred embodiment, the gene mutation perturbation pathways for constructing the first model are the Hedgehog signaling pathway, the ErbB signaling pathway and the Focal adhesion signaling pathway.

[0017] Furthermore, the calculation formula of the first model is as follows:

[0018]

[0019] Wherein, Hedgehog, ErbB, and Focal adhesion are the perturbation values ​​of Hedgehog signaling pathway, ErbB signaling pathway, and Focal adhesion signaling pathway, respectively.

[0020] As a preferred implementation, risk grading is performed based on the median of the total risk score in the training data.

[0021] As a preferred implementation, the calculation formula of the second model is as follows:

[0022]

[0023] In the formula, =Aspartate aminotransferase level; Child-Pugh A=0, B=1, C=1; Metastasis is distant metastasis, with metastasis=1, no metastasis=0. The hybrid model constructed by the present invention is verified by the validation set data, and the parameters are universal.

[0024] Another object of the present invention is to provide a computer program product, which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the steps of the above method are implemented.

[0025] The present invention uses an NGS panel targeting cancer-related genes and the R package "PMAPscore" to screen out key signaling pathways related to triple therapy for advanced liver cancer. In the training cohort, three signaling pathways were screened to establish the first model, and combined with clinical risk factors, a second model for assessing comprehensive risk was constructed. The risk score was then validated in a prospective internal cohort, an external immunotherapy cohort, and an external non-immunotherapy to determine the universality of the hybrid model. Compared with the prior art, the present application has the beneficial effect of being able to specifically and robustly divide patients with advanced HCC who are receiving triple therapy or PD-1 inhibitors into two different subgroups, each with a high or low risk of disease progression. The hybrid model has important clinical significance for clinical decision-making and personalized treatment of patients with advanced HCC. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of the present invention.

[0027] Figure 2 The PFS-related signaling pathways were integrated into GeneScore to evaluate the impact of gene mutations on the prognosis of patients with advanced HCC, including: (A) The relationship between all 123 signaling pathways and PFS in a retrospective HCC cohort treated with triple therapy using univariate Cox survival analysis; (B) Forest plot showing five signaling pathways associated with PFS (p < 0.1); (C) Forest plot showing three signaling pathways associated with PFS using multivariate Cox stepwise regression analysis (p < 0.1); (D) GeneScore distribution of patients with advanced HCC treated with triple therapy; (E) Heat map of perturbation scores of the three signaling pathways for each HCC sample: Blue: high perturbation score. White: low perturbation score.

[0028] Figure 3 It is the construction of a prognostic model for patients with advanced HCC treated with triple therapy.

[0029] Figure 4 Figure 2 is the Kaplan-Meier survival curve of PFS in the low GeneScore group and the high GeneScore group.

[0030] Figure 5 Evaluation and validation of the prognostic model for patients with advanced HCC treated with triple therapy, including (A) Kaplan-Meier survival curve based on OverallScore calculated by the mixed model, divided into high / low groups according to the median OverallScore; (B) Different responses of patients with advanced HCC with high / low OverallScore after triple therapy: CR, complete remission; PR, partial remission; SD, stable disease; PD, progressive disease; (C) Kaplan-Meier survival curves of 26 patients with advanced HCC in the validation set who received triple therapy, with OverallScore calculated using the second model and the same cutoff value as in the training set.

[0031] Figure 6 The figure is the evaluation and validation of the prognostic model for patients with advanced HCC treated with triple therapy, including (A) ROC curves of the hybrid model predicting 9-month, 18-month, and 24-month PFS; (B) AUC of the hybrid model predicting the change of PFS over time at different time points; (C) calibration graph of the hybrid model predicting 6-month, 9-month, and 12-month PFS rates. The predicted probability and actual probability of PFS are plotted on the x-axis and y-axis, respectively. The dotted line represents the ideal state graph, and the solid line represents the actual observed state graph.

[0032] Figure 7 This study explores the potential molecular mechanisms of triple therapy for the treatment of HCC patients from the perspective of biological processes and immune microenvironment; (A)-(C) Top 20 mutant genes in the low / high GeneScore group: (A) Gene mutation volcano plot; (B) CNTTB1 gene mutation frequency (C).

[0033] Figure 8 The present study explored the potential molecular mechanisms of triple therapy for HCC patients from the perspective of biological processes and immune microenvironment; (A) Gene set enrichment analysis (GSEA) of low GeneScore group and high GeneScore group, angiogenesis and hypoxia pathways were significantly enriched in the low GeneScore group; (B) Immunogram of cancer immune cycle in low / high GeneScore group, IGS1, whether there is T cell immune environment in the tumor, IGS2, tumor antigenicity; IGS3, initiation and activation; IGS4, transport and penetration; IGS5, tumor antigen recognition; IGS6 to IGS8, inhibitory factors prevent killing of cancer cells.

[0034] Fig. 9This study explores the potential molecular mechanisms of triple therapy for HCC patients from the perspective of biological processes and immune microenvironment; (A) GSEA comparison between low GeneScore group and high GeneScore group showed that antigen processing and presentation, BCR signaling pathway and chemokine gene sets were significantly enriched in the low GeneScore group; (B) Representative immunofluorescence images of CD3+ and CD4+ in tumor tissues of low / high GeneScore patients. DETAILED DESCRIPTION

[0035] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0036] Some of the tools and existing methods used in the examples are described as follows:

[0037] (1) Screening of signal pathways and fitting of the first model (Genescore computational model)

[0038] Univariate Cox regression analysis was performed using the survival analysis R package to screen signal pathways associated with prognosis. Multivariate Cox stepwise regression analysis was then used to screen candidate signal pathways that could be used to establish the first model (prognostic risk score Genescore calculation model).

[0039] (2) Gene set enrichment analysis and TME immune microenvironment analysis

[0040] Gene set enrichment analysis (GSEA) was performed using the clusterProfiler package. Gene sets for signaling pathways targeted by Lenvatinib, such as angiogenesis and hypoxia, were obtained from the MSigDB database, while immune-related gene lists were obtained from the Immport database (https: / / www.immport.org / shared / genelists). The abundance of tumor-infiltrating immune cells (TIICs) was quantified using single-sample GSEA in the GSVA package. Antitumor immune responses (cancer immunity cycle) were also assessed between different risk groups.

[0041] (3) Construction and verification of the second model (Overallscore calculation model)

[0042] Based on the gene risk score Genescore characteristic signal pathway and clinical factors, the Overallscore calculation model was constructed using the RMS software package. The prediction feasibility and reliability of the Overallscore calculation model were evaluated by calibration and ROC curve. Then, the clinical performance and net benefit of the Overallscore calculation model were evaluated by decision curve analysis (DCA).

[0043] (4) Immunofluorescence

[0044] Paraffin-embedded hepatocellular carcinoma tissue sections were immunofluorescently stained according to standard protocols. The immunofluorescence antibody was anti-CD3 antibody (Dako, Cat#A0452 IR503), and the nuclei were counterstained with 4'-6'-diamidino-2-phenylindole (DAPI). The labeled slides were scanned using the TissueFACS SL plus S system (TissueGnostics, Vienna, Austria; acquisition software: TissueFAXS SL V7.1.120) and analyzed by the advanced image analysis software StrataQuest (V7.1.1.129). Images from different channels were superimposed to visualize different labeling of the same tissue area. To identify tumor areas, the analysis software was used to automatically identify areas with obvious morphological features from tumor tissue. Positive and negative controls were included for each staining to ensure the reliability of the results.

[0045] (5) Statistical analysis

[0046] All statistical analyses were performed using R language version 4.1.1. The Wilcoxon test was used for comparisons between two continuous variables, and the Fisher exact tests were used for comparisons between categorical variables. Kaplan-Meier survival curves were constructed for prognostic analysis of categorical variables, and log-rank tests were used for statistical analysis. p < 0.05 was considered statistically significant.

[0047] Example 1

[0048] This example specifically illustrates a method for constructing a hybrid model for predicting PFS of patients with advanced HCC after triple immunotherapy.

[0049] A retrospective cohort of 82 patients with advanced HCC was used as the training set for model construction. All 82 patients received triple therapy of chemotherapy combined with anti-vascular drugs and immunotherapy. The enrolled patients mainly received HAIC combined with lenvatinib and tislelizumab. The detailed treatment plan is as follows: HAIC plan: oxaliplatin (130mg / m 2 , from 0 to 2 hours, day 1), folinic acid (200 mg / m 2 , from 2 hours to 4 hours, day 1), fluorouracil (400 mg / m 2 , 5-fluorouracil within 15 minutes, followed by 2400 mg / m 2Fluorouracil, 46 hours, day 1 and day 2). After HAIC administration, the catheter and sheath were removed. All patients received intravenous infusion of tislelizumab 101 and oral lenvatinib (12 mg / 60 kg; 8 mg / 60 kg or less) every three weeks. The method for obtaining tumor tissue and paired leukocytes is conventional technology and is not limited in this embodiment.

[0050] Secondly, formalin-fixed paraffin-embedded (FFPE) tumor tissue samples were used to detect somatic single nucleotide variations (SNVs) and insertion or deletion mutations. The library sequencing data of the patient's tumor tissue and paired white blood cell control were obtained, and the library data was used to construct a personalized tumor variation map of the patient. When detecting somatic single nucleotide variations (SNVs) and insertion / deletion (indel) mutations by FFPE tumor samples, the implementation of the present invention does not limit the specific method, and technicians in this field can directly use existing methods to complete it. For example, in this embodiment, black-PREP FFPEDNA kit (Analytik Jena, Germany) was used to separate DNA from FFPE fragments. The control group used Tiangen whole blood DNA kit (Tiangen) to extract DNA from peripheral blood lymphocytes, and whole blood (1600g) was centrifuged at room temperature for 10 minutes to obtain lymphocytes. Genomic DNA fragments (150-200 bp fragments) were segmented using a Covaris M220 focused ultrasound instrument (Covaris), and DNA libraries were constructed using the Illumina platform KAPA HTP library preparation kit (KAPA Biosystems). The DNA library (NimbleGen SeqCap EZ library; Roche) was then used to capture the major tumor-related gene sequences using a 769-gene panel and sequenced using an Illumina HiSeq X-Ten sequencer. All of the above were performed according to the manufacturer's instructions. Somatic SNVs were filtered and analyzed by VarScan2 (v2.4.2), and eligible variants had to meet the following filtering criteria: (i) sequencing coverage: control > 50x, tumor > 100x; (ii) mutant allele frequency > 2%; (iii) mutant allele read number > 2%; (iv) SNVs and Indels were located in exonic regions; (v) allele frequency in the Exome Integration Consortium (ExAC) database or the Genome Aggregation (gnomAD) database was < 0.5%. The final mutation list was used to calculate the PMPscore using the R package “PMAPscore”.

[0051] First, the PMAPscore package was used to calculate the signaling pathway perturbation score from the DNA mutation data of the baseline biopsy tissue of the retrospective cohort patients (who had not yet received triple therapy). Then, a single factor Cox regression analysis was performed on a total of 123 signaling pathways using the survival analysis R package ( Figure 2 A), and 5 signal pathways with a p-value less than 0.1 were screened out ( Figure 2 Furthermore, multivariate Cox stepwise regression analysis was performed on these five pathways, and finally three pathways were identified ( Figure 2 C), which are Hedgehog signaling pathway, ErbB signaling pathway and Focal adhesion signaling pathway, respectively, including 31 related genes, which were included in the final scoring system for predicting PFS ( Figure 2 In Figure C, a biomarker model Genescore based on signaling pathways was constructed. The patient's personalized mutation list was imported into the Genescore computational model to calculate the genetic risk score.

[0052] The calculation formula of the gene risk score Genescore is: Genescore =∑(Coefi ×Expi), where Coefi and Expi represent the risk coefficient and the characteristic signal pathway disturbance value, respectively, and the risk coefficient is obtained by fitting the retrospective cohort data.

[0053] The calculation formula in this embodiment is as follows:

[0054]

[0055] Wherein, Hedgehog, ErbB, and Focal adhesion are the perturbation values ​​of Hedgehog signaling pathway, ErbB signaling pathway, and Focal adhesion signaling pathway, respectively.

[0056] The patients were grouped according to Genescore values: Genescore ≤ 0: low risk; Genescore > 0: high risk.

[0057] According to the above calculation formula, each patient will receive his or her unique genetic risk score Genescore, and can be divided into high-risk patients and low-risk patients according to the cutoff value = 0 ( Figure 2 In the high-risk group, the perturbation scores of the Hedgehog signaling pathway and the Focal adhesion signaling pathway were higher. In contrast, in the low-risk group, the perturbation score of the ErbB signaling pathway was higher ( Figure 2 Middle E). At this cutoff value, the 30 high-risk patients (36.6%) in the training set had a shorter PFS than the 52 low-risk patients (63.4%) ( Figure 4 ). The results of the Genescore calculation model indicate that low-risk patients receiving triple therapy have a longer PFS than high-risk patients and may have a better prognosis.

[0058] Furthermore, clinical risk factors with significant association with PFS were screened out through univariate Cox regression analysis and multivariate Cox stepwise regression analysis. Three clinical characteristics, AST, Child-Pugh, and metastasis, were screened out. Combined with Genescore, a nomogram calculation model was constructed with a total of four risk factors to calculate the overall risk score Overallscore ( Figure 3 ), the calculation formula is as follows:

[0059]

[0060] In the formula, AST is the level of aspartate aminotransferase; Child-Pugh grading is grade A=0, grade B=1, grade C=1; Metastasis is distant metastasis, with the value of metastasis=1 and no metastasis=0.

[0061] The median Overallscore of 0.8 was used as the cutoff value: Overallscore ≤ 0.8: low risk; Overallscore > 0.8: high risk.

[0062] In order to enhance the convenience of clinical application, a computational tool for predicting PFS of HCC triple immunotherapy that is easy for clinicians to use can be developed. In this tool, the patient's AST, Child-Pugh, Metastasis, and mutant gene list are entered, and Genescore, Overallscore, and PFS prediction results can be automatically calculated.

[0063] Example 2

[0064] The overallscore model was validated based on a prospective internal validation cohort.

[0065] HCC patients were divided into high / low OverallScore groups according to the OverallScore cutoff value (=0.8). Survival analysis confirmed that the PFS of patients in the low OverallScore group was significantly better than that in the high score group (hazard ratio, 5; 95%CI, 2.66 ~ 9.38; p < 0.001; Figure 5 The objective response rate (ORR) of patients with low overall scores was 68.3%, which was higher than that of patients with high scores in the triple therapy HCC cohort (54%) ( Figure 5 (middle B).

[0066] Furthermore, 26 patients with advanced HCC who received triple therapy were prospectively recruited as the validation cohort to verify the reliability of the model. In the validation cohort, after triple therapy, the PFS of HCC patients in the low OverallScore group was superior to that in the high OverallScore group (HR, 2.8; 95%CI, 0.84 ~ 9.3; P = 0.046; P = 0.054). Figure 5 Middle C). The results showed that the OverallScore, which combines signaling pathway perturbations and clinical risk factors, could effectively predict PFS in patients with advanced HCC after HAIC combined with lenvatinib and tislelizumab treatment.

[0067] Example 3

[0068] The prediction accuracy of the hybrid model was verified by calculating the AUC and performing calibration.

[0069] The ROC results showed that the AUCs of the hybrid model in predicting 9-month, 18-month, and 24-month PFS were 0.863, 0.876, and 0.897, respectively. Figure 6 Middle A). The AUC for predicting PFS within two years was 0.846 ( Figure 6 The calibration graphs for predicting 6-, 9-, and 12-month PFS highly overlap with the ideal model and show good performance ( Figure 6 In addition, the DCA graph shows that the hybrid model has clinical practicality and high benefits in predicting 6-, 9-, and 12-month PFS, demonstrating its superior predictive ability in clinical application. Based on the hybrid model, HCC patients were divided into high-risk and low-risk groups, and survival analysis confirmed that the PFS of patients in the low-risk group was significantly longer than that of the high-risk group.

[0070] Example 4

[0071] Explore Genescore-related biological processes and gene signaling pathways

[0072] The present invention explores the biological processes and gene signaling pathways associated with the gene risk score Genescore from multiple levels. First, at the DNA mutation level, the gene mutation map ( Figure 7 Middle A) shows that the mutation frequencies of TP53 in the low GeneScore group and the high GeneScore group were 60.4% (32 / 53 cases) and 40% (12 / 30 cases), respectively. However, the mutation frequency of CTNNB1 in the low-score group was 3.8% (2 / 53 cases), much lower than the 53.3% (17 / 30 cases) in the high-score group; Figure 7 B and C).

[0073] Subsequently, gene enrichment analysis (GSEA) showed that the low GeneScore group had significantly enriched hallmark angiogenesis and hypoxia-related hallmark features ( Figure 8 Middle A). Lenvatinib, as a first-line treatment for unresectable HCC, targets multiple tyrosine kinases and inhibits angiogenesis. This may be one of the important reasons for the good prognosis of triple therapy in the treatment of low-risk advanced HCC. First, at the DNA mutation level, the volcano plot showing the odds ratio of all genes showed that the CNNTB1 mutation frequency in the triple therapy response group was significantly lower than that in the non-response group, while the mutation frequencies of other genes were basically the same. Subsequently, gene set enrichment analysis (GSEA) found that the marker features related to angiogenesis and hypoxia in the response group were significantly enriched. Analysis of the tumor's immune environment suggested that the tumor immunogenicity of the responder group was stronger.

[0074] Furthermore, analysis of the tumor's immune environment suggested that the low-risk group had stronger immunogenicity ( Figure 8 (middle B).

[0075] Further GESA analysis showed that antigen processing and presentation, BCR signaling pathway, and chemokine signaling pathway were significantly enriched in the responder group, indicating that the immune environment in the responder group was more active ( Fig. 9 Furthermore, immunofluorescence analysis confirmed that the CD3+ and CD4+ cell contents in the responder group were higher ( Fig. 9 In addition, the results of gene set variation analysis (GSVA) ​​of 28 immune subtypes showed that the content of memory effect CD4+, activated CD4+ and Th2 cells in the responder group was significantly higher than that in the non-responder group. This multi-faceted exploration revealed the biological processes and gene signaling pathways associated with risk scores.

Claims

1. A method for predicting progression-free survival of patients with advanced HCC after triple immunotherapy, characterized in that: include: Construct a personalized tumor mutation map based on the sequencing data of tumor tissues and paired leukocytes from patients with advanced HCC; Based on the tumor mutation information in the personalized tumor mutation map, the first model is used to calculate the genetic risk score; Then, the total risk score was calculated using the second model with genetic risk score and clinical variables as independent variables; Risk stratification was performed based on the cutoff value of the total risk score to predict the probability of PFS of patients receiving triple immunotherapy at different times; Among them, the first model and the second model are constructed with a retrospective cohort of patients with advanced HCC as training data, including: constructing a personalized tumor variation map based on tumor tissues of patients in the retrospective cohort before triple immunotherapy, calculating the perturbation level of tumor mutation information in the signal pathway by combining the cumulative effect and position effect of mutant genes, screening gene mutation perturbation pathways significantly associated with PFS, using the gene mutation perturbation pathways significantly associated with PFS as biomarkers, and constructing the first model based on regression analysis; The second model was established based on regression analysis with the calculation results of the first model and the clinical variables of the retrospective cohort patients as independent variables, wherein the clinical variables were clinical risk factors significantly associated with PFS.

2. The method according to claim 1, characterized in that The gene mutation perturbation pathway significantly associated with PFS was screened through univariate Cox regression analysis and multivariate Cox stepwise regression analysis, and the first model was established.

3. The method according to claim 1, characterized in that The calculation results of the first model and the clinical variables of the retrospective cohort patients were used as independent variables, and the second model was established based on the nomogram.

4. The method according to claim 1, characterized in that: Univariate Cox regression analysis and multivariate Cox stepwise regression analysis were used to screen out clinical risk factors that were significantly associated with PFS.

5. The method according to claim 4, characterized in that The clinical risk factors include aspartate aminotransferase level AST, Child-Pugh grade and metastatic status.

6. The method according to claim 1, characterized in that The gene mutation perturbation pathways used to construct the first model are the Hedgehog signaling pathway, the ErbB signaling pathway, and the Focal adhesion signaling pathway.

7. The method according to claim 6, characterized in that The calculation formula of the first model is as follows: ； Wherein, Hedgehog, ErbB, and Focal adhesion are the perturbation values ​​of Hedgehog signaling pathway, ErbB signaling pathway, and Focal adhesion signaling pathway, respectively.

8. The method according to claim 1, characterized in that: Risk stratification is performed based on the median of the total risk score in the training data.

9. The method according to claim 1, characterized in that: The calculation formula of the second model is as follows: ； In the formula, AST is the level of aspartate aminotransferase; Child-Pugh grading is grade A=0, grade B=1, grade C=1; Metastasis is distant metastasis, with the value of metastasis=1, no metastasis=0.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method described in any one of claims 1 to 9 are implemented.

Images