Background: The current study aimed to understand the genetic landscape and investigate the diagnostic and prognostic biomarkers of primary hepatocellular carcinoma (HCC). Methods: A cohort of 36 Chinese HCC samples with hepatitis B virus (HBV) infection was examined by whole-exome sequencing (WES). Prognosis-related alterations were identified and further verified in the TCGA database and GSE65372 profiles in the GEO database. A Chinese replication cohort of 180 HCC samples with HBV infection was collected to evaluate the candidate genes by immunohistochemical analysis. A receiver operating characteristic (ROC) curve analysis evaluated the prognostic power of candidate genes. Finally, EdU and transwell invasion assay were performed to detect the function of candidate genes. Results: A total of 11 novel genes showed a significant association with HCC in the discovery cohort. The data were verified using the GEO and TCGA databases, and the expression of ARID1A, CSMD1, and SENP was evaluated in the replication cohort. Furthermore, ARID1A, CSMD1, and SENP3 are effective prognostic biomarkers for HCC patients in the replication population. Conclusions: Molecular heterogeneity was detected in HCC patients, and ARID1A, CSMD1, and SENP3 were identified as effective HCC prognosis biomarkers. CSMD1 prevents HCC by suppressing cell invasion.