Using Integrated Multi-Omics Data Analysis to Identify 5-gene Signature for Predicting Survival of Patients with Hepatocellular Carcinoma

Abstract

Due to the poor prognosis for hepatocellular carcinoma (HCC) presently, a systemic analysis supported by the multi-omics data is extremely necessary to search for gene markers for the clinical prognostic prediction of HCC. The data on RNA-seq, single nucleotide polymorphism (SNP), and copy number variation (CNV), etc. were downloaded from TCGA, leading to a final of 367 samples, which were divided into training set and testing set randomly. In the training set, both prognosis-related genes and those with SNP or CNV were screened, which were incorporated for feature selection using the random forest method. The testing and GEO verification sets (N = 265) were used to verify the constructed gene-related prognosis model. qPCR was used to detect the expression of 5 genes in clinical specimens. After including genomic variant and prognosis-related genes, we got 78 candidate genes and 5 feature genes (CISH, LHPP, MGMT, PDRG1, and LCAT) eventually through random forest feature selection. The 5-gene signature is an independent prognostic risk factor for HCC patients. In addition, the signature shows good predicting performance and clinical practicality in train- ing set, testing set and external verification set. The results of qPCR based on clinical samples showed that the expression of PDRG1 was increased in colon cancer tissues and the expression of CISH, LHPP, MGMT and LCAT were decreased in colon cancer tissues. We identify the ability of 5-gene signature to serve as an innovative marker of survival prediction for patients with HCC.