An Integration Framework for Liver Cancer Subtype Classification and Survival Prediction Based on Multi-omics Data

Accurate prediction is helpful to the treatment of liver cancer. In this paper, we propose a method based on a combination of deep learning and network fusion to predict the survival subtype of liver cancer, of which Univariate Cox-PH regression model was used twice. We integrated RNA sequencing, miRNA sequencing, DNA methylation data and clinical data of liver cancer from TCGA to infer two survival subtypes. We then also constructed an XGBoost supervised classification model to predict the survival subtype of the new sample. Experimental results show that our model gives two subgroups with significant survival differences and Concordance index. We also use two additional confirmation cohorts downloaded from the GEO database to verify our multi-omics model. We found highly expressed stemness marker genes CD24, KRT19 and EPCAM and the tumor marker gene BIRC5 in two survival subgroups. Our method has great clinical significance for the prediction of HCC prognosis.