Liver cancer (LC) is the second leading cause of cancer-related deaths globally, yet the molecular mechanisms linking its progression with associated risk factors (RFs) remain poorly understood. To address this, we developed an integrative multi-stage framework combining bioinformatics, machine learning-based feature selection, survival modeling, and network analysis to identify robust biomarkers and pathways involved in LC progression. Unlike conventional biomarker discovery approaches, our strategy integrates multi-cohort transcriptomic and clinical datasets, enhancing robustness and reliability of findings. Initially, differentially expressed genes were identified from three Gene Expression Omnibus datasets for LC and its RFs. Next, using shared biomarkers, we constructed a gene-disease association (diseasome) network, revealing 230 unique genes, including 126 shared between LC and liver cirrhosis. Subsequently, RNA-seq and clinical data from The Cancer Genome Atlas (TCGA) were analyzed through combined and multivariate Cox survival models, identifying 70 prognostic genes. Among these, we identified RGS5, SULT1C2, CSM3, and CXCL14 as consistent survival-associated markers. Functional investigation of the 70 genes using enrichment and protein-protein interaction networks uncovered ten hub genes involved in key oncogenic pathways, including Oocyte meiosis, Lysine degradation and cell cycle regulation. These findings were further validated through literature and expression-level analysis. Additionally, an independent survival analysis using the full TCGA transcriptomic dataset identified 76 significant genes, with 18 overlapping the risk-associated gene set, reinforcing their prognostic value. Overall, this study demonstrates the potential of an integrative computational approach to uncover meaningful biomarkers and pathways in LC, offering valuable insights for future clinical and therapeutic strategies.