| Purpose:Studies on a genome-wide network in prognosis prediction of gastric cancer are rare.Previous studies have proposed and confirmed numerous biomarkers for gastric cancer.However,malignant tumors often involve multiple layers and different levels of genetic changes,including genome,transcriptome and proteome,or even epigenetic content.Selecting reasonable candidate factors from tens of thousands of biomarkers and comprehensively analyzing them as an independent feature can provide more comprehensive insights.Therefore,genetic networks containing a panel of abnormal factors from different regulatory levels may have a better prognostic value.This study aims to integrate various biomolecules(miRNAs,mRNA and DNA methylation)into a genome-wide network and develops a nomogram to predict overall survival of GC.Materials and methods:Level 3 data was downloaded from the TCGA database using TCGA-Assembler Module A in January 2019 and then was pretreated by Module B.The dataset included clinical data(age,sex,stage,primary site,grade,treatment and survival)and genome-wide data as follow:443 patients with clinical data;384 patients with expression levels of 1,871 miRNAs;377 patients with expression levels of 20,531 mRNA and 394 patients with 485,577 DNA methylation sites(Illumina methylation 450).Afterwards,an intersection with 332 samples in total among them was retained.Furthermore,patients missing active follow-up were excluded from the analysis,leaving 329 patients in the final cohort.Fourthly,genome-wide level 3 data whose expression level(miRNAs,mRNA and DNA methylation sites)was missing in more than 50%of all samples would be removed from final analysis.Finally,329 GC patients with 566 miRNAs,17963 mRNA and 396081 DNA methylation sites were chosen for further analysis.329 cases with GC are screened from TCGA as a training cohort and a random 150 examples is validation cohort.A genome-wide network is constructed based on batch Cox regression and LASSO.Nomogram is established to predict 1,3 and 5-year OS in training cohort,which is then assessed with calibration,discrimination and clinical usefulness in the validation cohort using ROC and DCA analysis of R program.GO and KEGG analysis are performed to explain the function of the genome-wide network.Results:A total of 329 GC patients in this study,212(64.4%)were man,and 117(35.6%)female,whose average age was 65.0±10.6years.In terms of pathological stage,38(11.6%)cases were identified as stage Ⅰ,117(35.6%)cases as stage Ⅱ,155(47.1%)cases as stage Ⅲ,and 19(5.8%)cases of stage Ⅳ.About treatment,303(92.1%)patients received surgery(280 cases of R0 surgery,14 R1 and 9 R2),and 146(44.4%)cases accepted Fluorouracil based chemotherapy.According to the median cutoff,165 samples were classified into a low-GS group(GS≤-0.137)and 164 samples a high-GS group(GS>-0.137)in the Genomic nomogram.A genomics score with 7 miRNAs,8 mRNA and 19 DNA methylation sites then divided into 9 models is a prognostic factor in multivariate analysis.Incorporating the genomics score into the genomics-based nomogram performs better in the estimation of OS than any single biomarker and the traditional TNM staging system,which could add prognostic value to the TNM staging.Multiple molecules and pathways are revealed to be related to the genome-wide network using DAVID 6.8.Conclusions:A genome-wide network is confirmed as a novel prognostic signature for GC,which could potentially play an important role in future clinical practice.A genome-wide network was developed and validated as a novel prognostic signature for gastric cancer through TCGA cohort.A combination of genomics score and TNM staging add prognostic value,proposing a more comprehensive subtyping system.The gene network could predict patient who benefit from chemotherapy to some degree as well.Then the biological function of this gene network is briefly described as well.New targets for gastric cancer may be detected from this gene network. |
PDF Full Text Request