Hepatocellular carcinoma (HCC) is an aggressive and heterogeneous liver cancer with restricted therapy selections and poor diagnosis. Although there have been great advances in genomics, the molecular mechanisms essential to HCC progression are not yet fully implicit, particularly at the single-cell stage. This research utilized single-cell RNA sequencing technology to evaluate transcriptional heterogeneity, immune cell infiltration, and potential therapeutic targets in HCC. A detailed bioinformatics pipeline used in the experiment included quality control, feature selection, dimensionality reduction using Principal Component Analysis (PCA), Uniform Manifold Approximation and Projection (UMAP), and t-distributed stochastic neighbor embedding (t-SNE), clustering, differential gene expression, pseudotime trajectory inference, and immune cell profiling with GSEA and survival analysis examining potential biomarkers of survival. Key findings include the identification of 1178 differentially expressed genes (DEGs), with macrophage infiltration contributing to immune evasion. Notably, APOE and ALB are linked to a better prognosis, while XIST and FTL are associated with poor survival. The potential drug candidates include IGMESINE in the case of SERPINA1 and PKR-A/MITZ for APOA2 in the gene-drug interaction analysis. Graph Neural Network (GNN) is used to predict drug-gene interactions and rank potential therapeutic candidates. The model shows robust predictive performance (R²: 0.9867, MSE: 0.0581) and identifies important drug candidates, such as Gadobenate Dimeglumine and Fluvastatin, and describes repurposing opportunities in network analysis, enhancing computational drug discovery for novel treatments. This research sheds new light on HCC tumor evolution, immune suppression, and the potential drug target based on the viewpoint of the importance of single-cell approaches in liver cancer research.