Construction Of Molecular Diagnostic Model Of Methylation And Integrated Analysis Of Multi-Omics Data Of Driver Genes In Esophageal Squamous Cell Carcinoma

Background and Objective:DNA methylation is the most extensively documented epigenetic modification with numerous essential functions in diverse diseases.Aberrant DNA methylation frequently occurs in the early tumor stages and is stable in various samples over time.It thus may be crucial biomarkers for clinical prediction of diagnosis in cancer.In China,esophageal squamous cell carcinoma(ESCC)is a prominent malignancy of the digestive system;however,the lack of effective methods for early diagnosis lead to the diagnosis of most patients in advanced stage or even with distal metastasis.Here,we reported the examination of genome-wide methylation profiles of 91 paired tumor and adjacent normal tissues in Chinese population,in order to clarify the molecular mechanisms of aberrant methylation changes in ESCC.We also aimed to identify potential DNA methylation biomarkers for diagnosis of ESCC.It would lay a foundation for the identification of epigenetic alterations associated with development of ESCC and potential targets.Methods:In the present study,Illumina 450K methylation chip was applied to detect the genome-wide DNA methylation levels in paired samples of 91 patients with ESCC.We compared methylation profiles of tumors with adjacent normal tissues in 91 patients.To elucidate the essential functions of aberrant DNA methylation in promoter or gene-body regions for ESCC,we leveraged the availability of RNA-sequencing data from the same individuals to compute the correlations between methylation and gene expression variation.We conducted the selection of methylation markers and the development of diagnostic model with Random-forest and LASSO methods.The diagnostic model was verified using the independent methylation datasets of ESCC from TCGA and GEO databases.We also evaluated the diagnostic ability of our classifier in multiple datasets by the area under ROC curves.Results:Here,we identified 35,577 differentially methylated CpG sites(DMCs)(43.46%hypermethylated vs 56.54%hypomethylated),through comprehensive analysis of genome-wide profiles of 91 paired tumor and adjacent normal tissues in Chinese population.Integration analysis revealed 134 genes whose expression was correlated with promoter DNA methylation,and a set of genes associated with gene-body methylation.Using a synthetical pipeline,we constructed a diagnostic model of eight-CpGs-based panel,strikingly,with high accuracy in training set,validation set and independent TCGA-ESCC cohort(AUC=0.99,0.98 and 0.96,respectively).We also integrated samples of esophagus mucosa from healthy individuals into the entire cohort,and the model can distinguish tumors clearly from adjacent tissues or normal esophagus mucosae(AUC=0.98).Additionally,our diagnostic model also could predict tumor versus non-tumor accurately in HNSC(AUC=0.97)and LUSC(AUC=0.96).Conclusions:We comprehensively characterized the genome-wide aberrant DNA methylation of ESCC by HumanMethylation450 BeadChip(450K array)from Illumina.It provided novel insights into the essential role of gene-body DNA methylation for ESCC.Our diagnostic model constructed in this study showed extremely high diagnostic accuracy in multiple methylation datasets,indicating its potential application in the early diagnosis of ESCC.We also found comparable performance of the classifier in HNSC and LUSC,suggesting the potential of this model in detection of other squamous cell carcinomas.Background and Objective:Multi-omics integration analysis was carried out on somatic mutation,copy number alteration,gene expression and DNA methylation of 20 driver genes in esophageal squamous cell carcinoma(ESCC),in order to fully elucidate the role and interaction of each alteration.Methods:Multigroup data of 95 cases of ESCC were downloaded from the cancer genome atlas(TCGA)and analyzed in combination with the existing 91 patients in our study.Patients were grouped according to the mutation of driver gene and the gene expression levels of groups were then compared with Student’s t-test.Spearman’s rank correlation was used to drive association analysis of copy number alteration,DNA methylation and gene expression.Kaplan-Meier survival curves of groups with or without driver gene mutation were compared by Log-rank test.Results:Multi-omics integration analysis revealed that,TP53,RBI,ZNF750 and PTCH1 showed significant difference between two groups(P=0.011,FC=1.43;P=0.045,FC=0.44;P=0.012,FC=2.20;P=0.011,FC=2.62);the expression levels of CUL3,PIK3CA,RBPJ,FBXW7,CDKN2A,PTEN,RBI and CERBBP were significantly correlated with copy number alteration;the expression levels of several driver genes were regulated by DNA methylation of promoter and gene body regions,and they were different between two datasets;the mutations of CREBBP(P=5.4E-05)and FAT1(P=0.0024)associated with ESCC prognosis and can serve as potential prognostic markers.Conclusions:Through the integration analysis of multi-omics data in ESCC,we discussed the correlations between somatic mutation,copy number alteration,DNA methylation and gene expression of 20 driver genes,as well as identified successfully two prognostic markers for ESCC.It also laid a foundation for further exploration of underlying mechanism of driver genes in ESCC.