Identification of Prognostic Genes in Hepatocellular Carcinoma

BACKGROUND: Previous studies have demonstrated the important role of tumor stem cells (TSCs) in the development of hepatocellular carcinoma (HCC); however, TSC-related genetic markers have not been investigated. AIM: The aim of the present study was to identify stem cell-related signature genes to predict the prognosis of HCC, using The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. METHODS: In total, 423 liver HCC tissue samples, including 373 tumor and 50 adjacent normal tissue samples from TCGA, and 115 primary tumor and 52 adjacent non-tumor tissue samples from the GEO GSE76427 database, were used in the present study. The non-negative matrix factorization (NMF) algorithm, t-distributed stochastic neighbor embedding (t-SNE) algorithm and Cox regression analysis were combined for model construction and validation. RESULTS: Overall, six clusters were identified using the NMF and t-SNE algorithms with 470 stem cell-related genes. The results demonstrated that patients in cluster 5 had the worst prognosis. For multivariate Cox survival analysis, 15 genes with optimal lambda values were chosen and eight genes were incorporated into the final regression model using the optimal Akaike information criterion value. Validation of the risk model using the aforementioned eight signature genes demonstrated the models strong reliability and stable predictive performance. CONCLUSION: The results of the present study indicated that the eight-gene (Hes family BHLH transcription factor 5, KIT ligand, methyltransferase-like 3, proteasome 26S subunit non-ATPase 1, Ras-related protein Rab-10, treacle ribosome biogenesis factor 1, YTH N6-methyladenosine RNA binding protein 2 and Zinc Finger CCCH-Type Containing 13) signature constructed by the model may be reliable in predicting the prognosis of patients with HCC.