Marker of CpG methylation site for early diagnosis of liver cancer

Abstract

The invention discloses a CpG methylation site marker for early diagnosis of liver cancer. And the number of CpG methylation sites is 43. In order to verify the precancerous methylation profile, a named five CpG-classifiers are established as biomarkers to develop and verify an HCC prediction line graph in an HBV infected patient retrospective queue (n = 604). We prove the practicability of the column chart, when the column chart is applied to screening who needs further diagnosis and preventive intervention, including chemotherapy or immunotherapy, of precancerous liver cancer patients.

Description

Technical Field

[0001] The present invention belongs to the technical field of cancer diagnosis, and in particular relates to a marker for early diagnosis of liver cancer using a CpG methylation site. Background Art

[0002] Hepatocellular carcinoma (HCC) is a cause of high morbidity and mortality, accounting for approximately 90% of all primary liver cancer cases. Patients who are positive for hepatitis B virus surface antigen (HBsAg) have a 10-25-fold increased risk of developing HCC compared to other uninfected patients. Early HCC screening is strongly recommended for HCC monitoring in high-risk hepatitis B (CHB) and liver cirrhosis (LC) patients. Screening patients for HCC that is difficult to diagnose in advance in the clinic is one of the biggest challenges. There is an urgent need for better biomarkers for blood analysis.

[0003] Abnormal methylation changes caused by HBV infection lead to HBV-related liver cancer and can serve as a good tumor marker. The pathogenesis of HCC is a dynamic process that includes changes in systemic molecular, cellular, and physiological states. Our cross-sectional study showed that methylation changes from CHB to HCC are associated with a critical transition period in the development of HCC. When the transition period occurs and what are the markers of the transition period are still unclear. Summary of the invention

[0004] The purpose of the present invention is to provide a CpG methylation site as a marker for early diagnosis of liver cancer.

[0005] A marker for early diagnosis of liver cancer, characterized in that the marker is 43 CpG sites.

[0006] The CpG sites are cg01088139, cg01116607, cg01236565, cg01296877, cg02997295, cg03570263, cg04088583, cg04552499, cg04588284, cg05331763, cg06656092, cg08138245, cg08603678, cg08958245, cg08965565, cg09445240, cg09983763, cg11823448, cg12131419, cg12482297, cg13556794, and cg148 one or more of 58817, cg15974568, cg16193207, cg16402757, cg16995742, cg17375863, cg18997918, cg19651115, cg19759847, cg20356637, cg21001065, cg22416262, cg23302570, cg23363182, cg23785275, cg24604274, cg25174338, cg25334461, cg25408960, cg26109791, cg26327119, and cg27284883.

[0007] Preferably, the CpG sites are cg13556794, cg17375863, cg25174338, cg11823448 and cg25334461.

[0008] A 5-CpG classifier for early diagnosis of liver cancer, wherein the risk score of the classifier=0.4283×cg13556794+2.7307×cg17375863-0.4972×cg25174338-0.4144×cg11823448-0.2322×cg25334461.

[0009] Among the 5-CpGs, 4 CpGs were significantly correlated with the time series, and 1 CpG was significantly associated with both pseudo-time and time series.

[0010] Patients with a risk score ≥ -0.15 are at high risk for early-stage liver cancer.

[0011] Beneficial effects of the present invention: To validate the methylation profile of precancerous lesions, the present invention established a named 5 CpG-classifier as a biomarker to develop and validate a HCC prediction nomogram in a retrospective cohort of HBV-infected patients (n=604). We demonstrated the utility of the nomogram when applied to screening precancerous HCC patients who require further diagnostic and preventive interventions, including chemotherapy or immunotherapy. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 Genes with CpG sites in their promoters were significantly correlated with pseudotime for pathway enrichment analysis.

[0013] Figure 2 Heat map of hierarchical clustering of 55 precancer and HCC stage samples for 1516 differentially methylated CpG sites.

[0014] Figure 3 A-A-Histogram of time before HCC onset based on the TimeAx model and the chronological time of 55 samples; B-Pseudo-time histogram; C-Comparative analysis of alpha-fetoprotein (AFP) between pseudo-time and time series; D-A-Comparative analysis of age between pseudo-time and chronological time.

[0015] Figure 4 A - Pseudotime based on the TimeAx model and the chronological time of the 55 samples; E - Correlation coefficients (Spearman correlation) of all 821835 CpG sites compared with the pseudotime or time before HCC onset; F - Volcano plot showing that CpG sites were not differentially methylated between the 8 precancerous stage samples (pseudotime ≥ 0.99) and the 25 precancerous stage samples (pseudotime < 0.99).

[0016] Figure 5 Compared with healthy controls, human autosomal imaging showed 5193 strongly significantly associated CpG sites within three to four years before HCC (A), one to two years before HCC (B), one year before HCC (C), and 0.5 months after HCC (D).

[0017] Figure 6 Figure 5 Quantitative distribution of methylation levels at specific CG sites and their correlation with the pathogenesis of HCC; 5193 robustly significantly associated CpG sites were evaluated using the TimeAx model; brown dots show the association of 5193 CpG sites with time (x-axis) and pseudotime (y-axis) (-log10 transformed, p-value based on linear regression), using a p-value threshold of 0.01.

[0018] Figure 7Figure 3 Heatmap of hierarchical clustering of 21 longitudinal PBMC samples from five patients using Spearman correlation of DNA methylation β-values ​​at 5193 robustly significantly correlated CpG positions (r>0.8; r<-0.8; p<6.8×10-8).

[0019] Figure 8 Figure 3. Heat map of hierarchical clustering of 17 paired precancer / HCC samples using DNA methylation β-values ​​of 5193 robustly significantly associated CG sites.

[0020] Fig. 9 To analyze the profiling and clustering of CpG site methylation levels using microarray significance mapping.

[0021] Fig.10 Heat map of hierarchical clustering of 21 longitudinal PBMC samples from five patients using DNA methylation β-values ​​at 43 robustly significantly associated CG sites (r > 0.8; r < -0.8; p < 6.8 × 10 -8 ).

[0022] Fig.11 Figure 2. Heat map of hierarchical clustering of 17 paired precancer / HCC samples using DNA methylation β-scores of 43 robust significantly associated CG sites.

[0023] Fig.12 Representative replicates of the 10-fold cross-validated LASSO coefficients for 43 candidate CpG sites.

[0024] Fig.13 Figure 2 shows the LASSO coefficient distribution of the coefficient paths of 43 CpG sites associated with HCC occurrence.

[0025] Fig.14 Flowchart for patient selection; CHB: chronic hepatitis B; LC: cirrhosis; HIV: human immunodeficiency virus.

[0026] Fig.15 To evaluate the five robust and significantly associated CpG sites using the TimeAx model; the association between pseudotime and time series, the five CpG sites (cg13556794, cg17375863, cg25174338, cg11823448, and cg25334461) are shown as heat maps (A) and scatter plots (BF).

[0027] Fig.16Figure 1. Nomogram for predicting the risk of HCC development within the next 12 months; (A) A nomogram was created to predict HCC incidence based on six independent prognostic factors, including age, sex, AFP, five CpG categories, first HBsAg-positive history, and PLR (platelet-to-lymphocyte ratio); (B) Area under the curve of the six independent predictors with 95% CI of the nomogram and HCC prediction over time since PBMC collection to HCC development (90-365 days); Time-dependent ROC curves of the nomogram for the probability of HCC development 6, 9, and 12 months after PBMC collection in the primary cohort (C) and the independent validation cohort (D).

[0028] Fig.17 The nomogram was used to predict the risk of HCC occurrence within the next 12 months; the calibration curves of the nomogram showing the consistency between the predicted and actual HCC occurrence probabilities 6, 9, and 12 months after PBMC collection in the primary cohort (E) and the independent validation cohort (F); the Kaplan-Meier curves showing the cumulative risk of HCC with 95% CI using the low-risk and high-risk groups defined by the nomogram in the primary cohort (G) and the independent validation cohort (H), and the log-rank test was used to determine the p value. DETAILED DESCRIPTION

[0029] In order to facilitate the understanding of the present invention, the present invention will be described more fully below. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0030] Example 1

[0031] 1. Experimental Methods

[0032] Patients and samples: All samples were obtained from patients treated at Beijing You'an Hospital, Capital Medical University. Detailed inclusion / exclusion criteria (Appendix E1) and baseline clinicopathological data, including demographics, past medical history, habits, routine blood tests, liver and kidney function tests, and other data, are described in the Supplementary Data (Appendix E2). The clinicopathological characteristics of the patients were collected at the time of PBMC sample collection. Whole blood specimens were collected before starting standard treatment for the patients. Fresh blood was collected into citrate and EDTA, and PBMCs were separated within 4 h by Ficoll-Hypaque density gradient centrifugation and stored in liquid nitrogen. Informed consent was obtained from all subjects, and the study was approved by the Institutional Ethics Committee of Beijing You'an Hospital.

[0033] A total of 22 HCC patients and their precancerous samples (n=55) and 18 healthy control samples were collected from June 13, 2010 to February 24, 2017. Seventeen paired precancerous / HCC samples (0-2 years before HCC vs 2 months after HCC) and 5 patients with 4 or 5 time points (3 within 4 years before HCC, 1 within 2 years before HCC, 1 year before HCC, and 1 within 0.5 months after HCC) were recruited.

[0034] To evaluate the prediction accuracy of HCC-related CpG sites, a retrospective collection of 604 patients with chronic hepatitis B and LC was performed with the help of the Beijing Hepatitis / AIDS Biobank from August 15, 2011 to August 28, 2018. According to the diagnosis time of CHB or LC, the enrolled patients were divided into a training cohort (n = 300, August 15, 2011 to April 14, 2015) and an independent validation cohort (n = 304, April 2015 to August 28, 2018) to construct and validate the HCC prediction nomogram. Excluding patients with incomplete baseline and follow-up information, the prediction process and sample size estimation of eligible patients were as follows Fig.14 and Appendix E3. Follow-up data were obtained by reviewing electronic medical records or telephone communications. The inventors calculated the follow-up time from the date of PBMC collection to the date of HCC diagnosis or August 27, 2019, whichever occurred first. The event was set as the occurrence of HCC within one year after PBMC collection.

[0035] Pathway analysis: Infinium Human Methylation EPIC Bead Chip Analysis Method by CapitalBio Technology according to the manufacturer's guidelines (Appendix E4) and multiple bisulfite sequencing (MBS) (Appendix E5). Based on the EPIC annotation file, all CpG sites were linked to the gene. The "promoter region" of a gene is closely linked to the negative regulation of the transcription process and consists of TSS1500, TSS200, 5'UTR or the first exon of the gene. The inventors selected differentially methylated genes whose "promoter region" contained differentially methylated CpG sites (DMC) for Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, www.QIAGEN.com / intelligence) and Kyoto Encyclopedia of Genes and Genomes (KEGG) to enrich related immune signaling and functional pathway analysis.

[0036] Statistical analysis: All analyses were performed using R language (version 4.2.2). The cutoff values ​​for quantitative variables were selected by applying the surve_cutpoint function implemented in the “surveminer” package or the reference range of the test items. Infinium EPIC beadchip arrays were analyzed using the Chip Analysis Methylation Pipeline (ChAMP). Bonferroni-corrected p values ​​< 0.05 were considered significant. TimeAx, an algorithm based on longitudinal molecular data, predicted HCC in pseudo-temporal locations for all samples. Microarray Significant Profiles (maSigPro) was used to screen CpG sites in significant methylation profiles that differed over the time course.

[0037] Methods: A Cox proportional hazards model based on the least absolute shrinkage and selection operator (LASSO) was used to construct a nomogram. Stratified analysis was used to estimate the relationship between the 5-CpG classifier and HCC in the next year within different subsets of confounding variables. Time-dependent receiver operating characteristic (ROC) curves were obtained using the "survivalROC" package. The nomogram was developed using the "rms" package. The area under the ROC curve (AUC) was used to assess the predictive ability. Validation was performed using a bootstrap method with 1000 resamplings, using bias-corrected calibration. Extension packages including "ggplot2", "surfminer", and "pec" were also used.

[0038] Appendix E1: Inclusion and Exclusion Criteria Inclusion and exclusion criteria were applied to this study and were as follows: Inclusion criteria for chronic hepatitis B were hepatitis B surface antigen (HBsAg) seropositivity for 6 months or longer according to the Asia Pacific Clinical Practice Guidelines for the Management of Hepatitis B: 2015 Update. Patients with hepatitis B-related cirrhosis (LC) were diagnosed according to the Guidelines for the Prevention and Treatment of Chronic Hepatitis B (2010 edition) based on a combination of clinical, laboratory, and imaging features and liver biopsy. The diagnosis of HCC was based on the radiological / histological criteria of the 2012 EASL Clinical Practice Guidelines: EASL Clinical Practice Guidelines: Management of Hepatocellular Carcinoma. HCC was classified according to the Barcelona Clinical Cancer (BCLC) staging system. Exclusion criteria were as follows: age less than 18 years; other types of cancer; co-infection with human immunodeficiency virus, hepatitis C virus, or hepatitis D virus; liver damage caused by drug intake coexisting with autoimmune hepatitis; pregnancy and lactation.

[0039] Appendix E2: Demographic and clinicopathological data: In this study, the demographic and baseline clinicopathological data were as follows: (1) Demographic data: age, gender; (2) Past medical history: history of first HBsAg positivity, history of CHB, LC or DLC, history of antiviral treatment; (3) Habits: history of smoking and drinking; (4) Routine blood tests: white blood cells, neutrophils, lymphocytes, monocytes, hemoglobin, platelets; (5) Liver and kidney function tests: Child-Pugh grade; alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), direct bilirubin (DBIL), total protein (TP), albumin, globulin, γ-glutamyl transpeptidase (γ-GT), alkaline phosphatase (ALP), prealbumin, total bile acid (TBA); (6) Other indicators: alpha-fetoprotein (AFP) and viral load.

[0040] Appendix E3: Sample size estimation, in the inventor's study, according to the time of diagnosis of HBVLD, the selected patients were divided into a primary cohort (n = 300, August 15, 2011 to April 14, 2015) and an independent validation cohort (n = 304, April 2015 to August 28, 2018) to construct and validate the HCC screening nomogram. According to the calculations performed in PASS version 11.0.7, the minimum sample size for constructing the nomogram was 270 (the hazard ratio of detecting the combination of HCC-related CpG sites at a significance level of 0.05 was equal to the 85% power of 3). Since the multivariate regression of the combination of HCC-related CpG sites on other covariates in Cox regression was expected to have an R square of 0.15, the sample size was adjusted. In our study, data from 300 patients in the training cohort were used to construct the nomogram, which was able to meet the sample size requirements.

[0041] For the validation cohort, the inventors used MedCalc to calculate the sample size required to compare the AUC of HCC incidence within one year (about 0.75) with the null hypothesis value (0.6). Type I error (α level, significant, bilateral) selected 0.05, and type II error (β level) selected 0.10 (power of 90%). The sample size ratio of the negative / positive group was 0.111. Finally, 31 cases were needed in the positive group and 273 cases in the negative group, for a total of 304 cases.

[0042] Appendix E4: Infinium Human Methylation EPIC Chip Analysis: PBMC DNA was treated with bisulfite to convert non-methylated cytosine residues to uracil (ZYMO Research, Irvine, USA) and tested using the Infinium HumanMethylation EPIC BeadChip array (Illumina, Inc., San Diego, CA) from CapitalBio Technology according to the manufacturer's guidelines. Briefly, the converted genomic DNA was isothermally amplified, digested, precipitated, denatured, and hybridized on the BeadChips. The BeadChips were then washed and extended with biotin-modified ddCTP, ddGTP or DNP-modified ddATP, ddTTP before scanning with the iScan system (Illumina, Inc., San Diego, CA). Thus, 866,836 CpG sites were measured on the EPIC array. The DNA methylation raw data were preprocessed using the Chip Analysis Methylation Pipeline (ChAMP). It mainly includes loading raw intensity data (IDAT) files, filtering probes, quality checks, normalization, and batch effect correction. Probes were removed as follows: detection P value > 0.01, bead number < 3 in at least 5% of samples, containing single nucleotide polymorphisms (SNPs) aligned to multiple positions, located on chromosomes X, Y, and non-CpG sites.

[0043] Appendix E5: Multiple bisulfite sequencing (MBS): MBS was used to measure the methylation levels of 43 HCC-related CpG sites in 604 PBMC samples. PBMC DNA was converted (ZYMO Research, Irvine, USA), and the concentration of single-stranded DNA was measured using Qubit 2.0 (Thermo, Carlsbad, CA, USA). A set of 43 pairs of “no CpG” primers were designed to target all candidate CpGs. Library preparation was performed using nested polymerase chain reaction, and bisulfite-converted DNA was bound as a template by 43 pairs of primers in the first round of polymerase chain reaction. In the second PCR reaction, the first-round PCR amplicon was used as a template, and the barcoded P7 universal primer and the barcoded P5 universal primer were used as the forward primer and reverse primer, respectively. Then, the final PCR amplicon was purified using AMPure XP beads. These libraries were then quantified and pooled. Paired-end sequencing of the libraries was performed on a HiSeq XTen sequencer (Illumina, San Diego, CA). MBS raw data were processed by SangonBiotech. Cutadapt (version 1.2.1) was applied to remove adapter sequences; PRINSEQ lite (v 0.20.3) was used to remove low-quality bases (Q < 20) from 3' to 5' of the reads; and Bismark (version v0.22.1) was applied to CpG site detection using default parameters.

[0044] 2. Experimental Results

[0045] (I) Comparative analysis of HCC trajectory pseudo-time and chronological time

[0046] Table 1 shows the clinical and pathological characteristics of 22 HCC patients and their precancerous samples collected at different time points. The TimeAx model predicts the specific pseudotime ( Figure 3 The inventors observed a high correlation between the predicted pseudotime and the time before HCC onset (or chronological time) (r = 0.69, Figure 3 Meanwhile, both types of time trajectories were highly correlated with 35 clinical parameters (r = 0.77). Pseudo-time was more highly correlated with AFP level and age than chronological time, which were important independent factors associated with HCC ( Figure 3 Middle C and Figure 3 (middle D).

[0047] Table 1 Stratified analysis of clinicopathological risk factors for 5-CpG-classifier

[0048]

[0049]

[0050] Note: PLR is the platelet-to-lymphocyte ratio.

[0051] Next, we evaluated CpG site methylation levels as a function of pseudotime. Differences in correlations (pseudotime vs. chronological time) were observed, while consistency between association patterns was maintained (r = 0.75; Figure 4 (E). Further analysis showed that CpG methylation levels were more strongly correlated with time. More than 26.4% of CpG sites were correlated with time alone, while the correlation with pseudotime or both measurements was only 8% and 5.7%, respectively. Some CpG sites that were positively or negatively correlated with pseudotime were located within the “promoter region” (TSS1500, TSS200, 5'UTR, and the first exon of the gene) of immune checkpoint genes (ICGs). These CpG sites mostly had negative regulatory effects on the transcription of ICGs, including PD-1, CTLA-4, LAG3, TIM3, BTLA, B7H3, VISTA, TIGIT, CD226, and CD96. The methylation changes of ICGs from precancerous to cancerous stages could be captured by pseudotime. KEGG pathway enrichment analysis was performed only on differentially methylated genes whose “promoter region” was occupied by DMC. The results showed that chemokine signaling pathway, T cell receptor signaling pathway, PD-L1 expression and PD-1 checkpoint pathway, and Th17 cell differentiation were strongly correlated with the pseudotime of HCC ( Figure 1 ).

[0052] Using a disease pseudotime cutoff of 0.99, the eight precancerous stage samples were divided into the HCC group. 1516 DMCs (Bonferroni-corrected p value < 0.05; |deltaβ|> 0.05) were analyzed for differences between the eight precancerous stage samples and HCC samples using Limma software. These DMCs did not show good discrimination ability between precancerous stage and HCC stage samples ( Figure 2 ). Meanwhile, there was no DMC between the 8 precancerous stage samples (pseudo-time ≥ 0.99) and the 25 precancerous stage samples (pseudo-time < 0.99) ( Figure 4 Middle F).

[0053] These results suggest that TimeAx dynamic modeling is able to capture the immune checkpoint signaling characteristics of HCC trajectory. However, more specific information and details can be obtained from the time series of HCC development.

[0054] (II) Time series analysis of HCC trajectories showed that the year before HCC was a critical period of transition

[0055] To further explore the methylation changes in the peripheral immune system before HCC development, 21 PBMC samples from five patients were collected at four or five time points from the LC stage to the HCC stage. Spearman correlation analysis was performed between the days before and after the HCC diagnosis date, and the beta values ​​of all CpG sites for each sample were formed. This analysis revealed a broad signature of DNA methylation associated with the pathogenesis of HCC. A genome-wide view of DNA methylation changes at 5193 strongly significantly correlated CpG sites (|r|>0.8; p<6.8×10-8) is shown in Figure 3. Figure 5 From the TimeAx model, the inventors determined that 63.65%, 0.39%, and 30.85% of the 5193 CpG sites were significantly associated with time, pseudotime, or both, respectively ( Figure 6 ). Hierarchical cluster analysis showed that precancerous samples (>12 months) clustered differently from all individual HCC samples and precancerous samples (≤12 months) ( Figure 7 ). The features of 5193 CpG sites were validated using 17 paired precancerous / HCC samples. Hierarchical cluster analysis showed clustering of all individual HCC patients and their precancerous stage (≤12 months) versus precancerous stage (>12 months) ( Figure 8 ). The above results indicate that the DNA methylation profiles of PBMCs in these precancerous stages (≤12 months) have parallel features with early HCC, which provides a key opportunity for the pathogenesis and diagnosis of early HCC. Comparative IPA canonical pathway analysis showed that the PD-1 / PD-L1 pathway in CD4+ or CD8+ T lymphocytes was activated in the 12 months precancerous and HCC stages. Overall, DNA methylation changes in PBMCs show features of advanced immunomodulatory functions and are consistent with a new role for the immune system in HCC.

[0056] MaSigPro divided 5193 significantly correlated CpG sites into 9 clusters based on their similar methylation features ( Fig. 9). The 9 clusters were further divided into “upper and lower” (clusters 1, 2, 6, and 7) and “lower and upper” (clusters 3, 4, 5, 8, and 9) according to the pattern, corresponding to the change from precancerous stage to HCC stage. The inventors selected 43 candidate CpG sites from clusters 2, 5, 7, and 9 to predict HCC in the next 12 months based on the maximum difference between HCC patients in the precancerous stage (≤12 months) and precancerous stage (>12 months) (CpG sites are cg01088139, cg01116607, cg01236565, cg01296877, cg02997295 , cg03570263, cg04088583, cg04552499, cg04588284, cg05331763, cg06656092, cg0 8138245, cg08603678, cg08958245, cg08965565, cg09445240, cg09983763, cg11823 448, cg12131419, cg12482297, cg13556794, cg14858817, cg15974568, cg16193207, cg16402757, cg16995742, cg17375863, cg18997918, cg19651115, cg19759847, cg20 Using hierarchical clustering, the 43 candidate CpG sites successfully separated the precancerous samples (>12 months) from all individual HCC samples and the precancerous samples (≤12 months) into two discrete groups in the 21 longitudinal samples ( Fig.10 ) and 17 paired precancerous / HCC samples ( Fig.11 ).

[0057] (III) Construction of a 5-CpG classifier for predicting HCC in the next year

[0058] The inventors further validated the predictive value of 43 HCC-associated CpG sites in a cohort of HBV-infected patients. A total of 1456 HBV-infected patients had PBMC samples available, of which 633 had complete baseline and follow-up information. Excluding other types of cancer, HIV co-infection, and death, this study ultimately recruited 604 eligible patients ( Fig.14In this retrospective study with a follow-up of 6.8 years (mean, 2.3 years), 67 patients developed HCC within the next year.

[0059] Using MBS data, the inventors applied LASSOCox regression to construct a predictive 5-CpG classifier ( Fig.12 and Fig.13 ), where risk score = 0.4283 × cg13556794 + 2.7307 ×

[0060] cg17375863-0.4972×cg25174338-0.4144×cg11823448-0.2322×cg25334461. Four CpGs were significantly associated with time series, and one CpG was significantly associated with both pseudo-time and time series ( Fig.15 ). The inventors calculated the HCC risk score for each patient. Patients with a risk score ≥ -0.15 are at high risk for HCC. In the primary cohort, the one-year cumulative risk of HCC in low-risk and high-risk patients was 5.0% (0.1-9.7%) and 12.4% (7.50%-17.0%), respectively (HR=2.95 (1.46-5.95), p=0.0026), and in the independent validation cohort it was 5.9% (0.70%-10.80%) and 18.5 (12.7-23.8) (HR=3.21 (1.26-8.19), p=0.0147).

[0061] (IV) Development and validation of a nomogram for early HCC prediction based on the 5-CpG classifier

[0062] To provide clinicians with a quantitative method to predict the probability of HCC in HBV-infected patients within the next year, a nomogram was constructed ( Fig.16 A). The coefficients of the six factors of the HCC prediction nomogram are listed in Table 2. The 5-CpG classifier has excellent prediction accuracy and a major contribution to the prediction performance ( Fig.16 B). In the main queue ( Fig.16 C) and an independent validation cohort ( Fig.16 In D), the time-dependent ROC curve shows the AUC values ​​for predicting HCC at 6, 9, and 12 months since PBMC collection. The calibration plot shows that the nomogram is consistent with the primary cohort ( Fig.17 E) and an independent validation cohort ( Fig.17 F) were very consistent with the actual HCC incidence rate. Using this nomogram, the one-year cumulative risk of HCC in the low-risk and high-risk groups in the primary cohort was 3.0% (0.1-5.8%) and 17.90% (11.00%-24.3%), respectively (p<0.0001) ( Fig.17G), 4.5% (1.20%-7.80%) and 27.3 (18.90-34.90) in the independent validation cohort (p>0.0001) ( Fig.17 H). These results suggest that the early HCC prediction nomogram is helpful in identifying the precancerous stages of HCC patients.

[0063] Table 2 Coefficients of HCC model Cox regression was used to construct the model

[0064]

[0065] Note: SE = standard error; HR = hazard ratio; CI = confidence interval, PLR = platelet-to-lymphocyte ratio

[0066] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A marker for early diagnosis of liver cancer, characterized in that: The markers are 43 CpG sites.

2. The marker for early diagnosis of liver cancer according to claim 1, characterized in that: The CpG sites are cg01088139, cg01116607, cg01236565, cg01296877, cg02997295, cg03570263, cg04088583, cg04552499, cg04588284, cg05331763, cg06656092, cg08138245, cg08603678, cg08958245, cg08965565, cg09445240, cg09983763, cg11823448, cg12131419, cg12482297, cg13556794, and cg148 one or more of 58817, cg15974568, cg16193207, cg16402757, cg16995742, cg17375863, cg18997918, cg19651115, cg19759847, cg20356637, cg21001065, cg22416262, cg23302570, cg23363182, cg23785275, cg24604274, cg25174338, cg25334461, cg25408960, cg26109791, cg26327119, and cg27284883.

3. The marker for early diagnosis of liver cancer according to claim 2, characterized in that: The CpG sites are cg13556794, cg17375863, cg25174338, cg11823448 and cg25334461.

4. A 5-CpG classifier for early diagnosis of liver cancer, characterized in that: The risk score of the classifier = 0.4283×cg13556794+2.7307×cg17375863-0.4972×cg25174338-0.4144×cg11823448-0.2322×cg25334461.

5. The 5-CpG classifier for early diagnosis of liver cancer according to claim 4, characterized in that: Among the 5-CpGs, 4 CpGs were significantly correlated with the time series, and 1 CpG was significantly associated with both pseudo-time and time series.

6. The 5-CpG classifier for early diagnosis of liver cancer according to claim 4, characterized in that: Patients with a risk score ≥ -0.15 are at high risk for early-stage liver cancer.

Images