Studies On Early Diagnostic And Surveillance Models Of Hepatocellular Carcinoma Based On Whole-genome Sequencing Of CfDNA

BACKGROUNDS&AIMSPrimary liver cancer(PLC)is the fourth leading cause of cancer-related death worldwide.Hepatocellular carcinoma(HCC)is the most common histological subtype of PLC,which accounts for approximately 90% of PLC cases.More than 90% of patients with HCC have the background of chronic liver disease.The most important risk factor in HCC development is liver cirrhosis(LC)caused by various etiologies.HCC high-risk populations of male over 35 years old and female over 45 years old are recommended to be under surveillance by abdominal ultrasound(with or without AFP)twice a year according to the guidelines of Chinese society of Clinical Oncology(CSCO)on diagnosis and treatment of HCC.Detected nodules should be further diagnosed by CT / MRI or pathological biopsy.Liquid biopsy is a new technology for noninvasive detection and dynamic analysis of a variety of biological samples,which exhibits the advantages such as simple testing procedure,good reproducibility,and accurate reflection of biological information of the original or metastatic tumor.Liquid biopsy not only has great potential to be applied to early diagnosis,prognosis evaluation,dynamic monitoring of tumor progression and therapeutic assessment,but also helps to understand the mechanism of tumor development and discover new targets of drugs.Cell-free DNA(cf DNA),one of the key biological markers of liquid biopsy,has been proved to carry tissue-of-origin information.In particular,the methylation and 5-hydroxymethylation of cf DNA have been confirmed as novel tumor diagnostic and screening biomarkers.This project aims to find new HCC-specific serological biomarker panels and provide novel diagnostic and surveillance strategies for early HCC through whole-genome analysis of cf DNA.METHODS1.A total of 5561 participants including HCC patients,LC patients,and healthy controls were enrolled from 16 hospitals across 11 provinces of China.The enrolled patients were followed up twice a year for three years.10 m L of peripheral blood samples were collected from each individual in the cohort at each follow-up after enrollment.cf DNA was extracted from plasma.2.DNA hydroxymethylation sequencing(5hm C-seq),low-pass whole genome sequencing(WGS),and Circulating Single-Molecule Amplification and Resequencing Technology(c SMART)were used to obtain the genomic characteristics of cf DNA such as DNA hydroxymethylation,copy number variation(CNV),nucleosome footprint(NF),5′end motif,DNA fragment size,and single-base substitution.Compare the differences of cf DNA in HCC,LC and healthy control populations.3.Analyze the cf DNA data of enrolled populations.Use machine learning methods such as logistic regression,Least Absolute Shrinkage and Selection Operator(LASSO)regression,and random forest algorithm to screen out the unique characteristics of HCC,and detection models of HCC were preliminarily constructed in a retrospective study.4.Evaluate the performance of the models by plotting the receiver operating characteristic(ROC)curve and calculating the area under curve(AUC),sensitivity,specificity,accuracy,positive predictive value(PPV),negative predictive value(NPV)of the models.5.Validate the value of the model in a real-world cross-sectional study.Compare the performance of liquid biopsy model and traditional methods in HCC detection.Clarify the advantages of cf DNA-based model in detection of HCC(especially early and very early HCC).6.Focus on the key molecules screened out during the analysis process,and further study their potential functions through molecular biology experiments.7.Use the prospective data to analyze the potential of the simple blood test for predicting risk of developing HCC.Regular follow-up was carried out on the false-positive LC patients who were misdiagnosed as HCC in the blood test,the dynamic changes of cf DNA and serological marker AFP were monitored,and ultrasound and CT / MRI were also performed.Compare the development of HCC in cf DNA false-positive LC patients and cf DNA negative LC patients within three years,so as to explore whether the cf DNA-based surveillance model can detect early HCC earlier than the current screening methods.RESULTS1.The low-pass WGS and 5hm C-seq data showed that lower sequencing depth at the transcription start sites(TSS),slightly lower overall read count of cf DNA with 5hm C,more viable and much shorter fragment size were found in HCC compared to LC or healthy group.Additionally,the pattern of base mismatch is similar between HCC and LC groups but significantly different between HCC cases and healthy control.2.Non-cirrhotic HCC exhibits a distinct NF distribution pattern.The comprehensive analysis of cf DNA NF data,WGS data of HCC tumor tissue,data from public databases,and HCC mouse model suggests that MID1IP1 may play an important role in HCC development.3.By c SMART analysis,mutation of 931 gene sites in 1185 HCC and LC samples were detected.We drew the distinct cf DNA mutation landscapes of HCC and LC patients,and screened out gene sites with significantly different mutation rates between HCC and LC groups.4.Through combination of three HCC-specific mutation sites and AFP,AFP-L3 and PIVKA-II,we developed a novel HCC diagnostic model named “Combined method” with the advantages of high sensitivity,high accuracy,and simple detection process.5.A surveillance model for HCC containing five features of cf DNA including 5hm C、5’ end motif、cf DNA fragment size,NF and CNV named “Pre Car Score” was constructed through analysis of cf DNA characteristics of 482 HCC and 2074 LC cases using low-pass WGS and 5hm C-seq through logistic regression and LASSO regression.6.We conducted a multi-center,cross-sectional study to validate the performance of Pre Car Score in a large population at high risk of HCC.The application of Pre Car Score alone achieved 1.2 or 3.9 folds increase in sensitivity for early and very early HCC surveillance compared to US or AFP method.In addition,among US-negative,AFP-negative and PIVKA-II-negative early HCC cases,Pre Car Score still exhibited high sensitivity for HCC surveillance,which showed Pre Car Score as a powerful supplement to the current recommended surveillance methods.7.Through follow up the LC patients with false-positive Pre Car Score for three years,we found that Pre Car Score-positive LC patients are more prone to HCC than those with negative Pre Car Score,suggesting that Pre Car Score has the potential to predict risk of HCC.CONCLUSIONSWe performed a multi-center,cross-sectional study to explore the unique characteristics of cf DNA in HCC patients,and developed a cf DNA-based diagnostic model for HCC.The diagnostic performance of the novel model was found superior to the current serological biomarkers.We also built Pre Car Score for HCC surveillance,which exhibited an excellent performance for early and very early HCC detection.Pre Car Score showed itself as a powerful supplement to current surveillance methods for HCC including ultrasound and AFP.Additionally,we prospectively verified the predictive ability in HCC development of Pre Car Score in some high-risk populations.In conclusion,this research aims to investigate the feasibility of cf DNA as a tumor biomarker for HCC and find novel strategies for early diagnosis and surveillance of HCC.In conclusion,this study established cf DNA-based multi-marker diagnostic and surveillance models for HCC,providing new methods and strategies for the early diagnosis and intervention of HCC.