Lysosomes play an important role in the pathological processes of cancer development. However, its effects on the prognosis and tumor microenvironment of hepatocellular carcinoma (HCC) remain unclear. Therefore, we aim to explore the novel molecular subtypes of HCC via lysosome-related genes (LRGs) for prognosis and therapy prediction in this study.Using the data of TCGA, differential expression and survival analyses were performed. Consequently, 109 key LRGs were obtained and 374 HCC samples were clustered into two groups: C1 and C2. A three-gene prognostic prediction nomogram was constructed using WCGNA and Cox regression analyses. Furthermore, pathway enrichment conditions, immune infiltration, immune checkpoint expression, and drug sensitivity were analyzed for the two subtypes. RT-qPCR was also used to validate the expression of the selected key LRGs.Key LRGs were highly expressed in the C1 subtype, and their prognosis was worse. The degree of immune cell infiltration and pathway enrichment results were also significantly different between the two subtypes. Furthermore, the three-gene prognostic prediction nomogram including LAPTM4B, PRKCD and LPCAT1, had a relatively high prognostic prediction ability. Meanwhile, the expression of immune checkpoints, human leukocyte antigen, and TIDE score were higher in the C1 subtype, suggesting that immune evasion was more likely to occur in this subtype. Drug sensitivity analysis showed that several drugs were more sensitive to C1 subtypes and might serve as drug candidates for these patients.We identified two novel molecular subtypes of HCC based on LRGs, and found that the LRGs related subtypes demonstrated significant efficacy in predicting the prognosis and therapeutic outcomes for patients with HCC. Moreover, a novel prognostic prediction nomogram was also developed, which possessed excellent prognostic prediction capabilities. We hope the novel LRG-related subtypes and nomogram of HCC would provide new insights and guidelines for clinical practice in the future.© 2025 Yao et al.