LAUSR

Background The initiation and progression of tumors were due to variations of gene sets rather than individual genes. This study aimed to identify novel biomarkers based on gene set variation analysis (GSVA) in hepatocellular carcinoma. Methods The activities of 50 hallmark pathways were scored in three microarray datasets with paired samples with GSVA, and differential analysis was performed with the limma R package. Unsupervised clustering was conducted to determine subtypes with the ConsensusClusterPlus R package in the TCGA-LIHC (n = 329) and LIRI-JP (n = 232) cohorts. Differentially expressed genes among subtypes were identified as initial variables. Then, we used TCGA-LIHC as the training set and LIRI-JP as the validation set. A six-gene model calculating the risk scores of patients was integrated with the least absolute shrinkage and selection operator (LASSO) and stepwise regression analyses. Kaplan–Meier (KM) and receiver operating characteristic (ROC) curves were performed to assess predictive performances. Multivariate Cox regression analyses were implemented to select independent prognostic factors, and a prognostic nomogram was integrated. Moreover, the diagnostic values of six genes were explored with the ROC curves and immunohistochemistry. Results Patients could be separated into two subtypes with different prognoses in both cohorts based on the identified differential hallmark pathways. Six prognostic genes (ASF1A, CENPA, LDHA, PSMB2, SRPRB, UCK2) were included in the risk score signature, which was demonstrated to be an independent prognostic factor. A nomogram including 540 patients was further integrated and well-calibrated. ROC analyses in the five cohorts and immunohistochemistry experiments in solid tissues indicated that CENPA and UCK2 exhibited high and robust diagnostic values. Conclusions Our study explored a promising prognostic nomogram and diagnostic biomarkers in hepatocellular carcinoma.