AI-integrated single-cell multi-omics decodes the hepatocellular carcinoma metabolism-immune axis: a new strategy for precision therapeutic targeting.

Background: Immune checkpoint inhibitors have improved outcomes for hepatocellular carcinoma, yet most patients do not respond because the tumor's metabolic environment suppresses immune cells. Single-cell RNA sequencing has revealed extensive immune diversity, but conventional analyses cannot link cell states to their physical location or to the metabolic signals that drive dysfunction. Methods: In this review, we examine how artificial intelligence, combined with single-cell and spatial multi-omics, can decode the metabolism-immunity network in liver cancer. We highlight two key metabolic switches: lipid uptake through a scavenger receptor that triggers ferroptosis in killer T cells, and lactate-induced changes in gene regulation that lock macrophages into a tumor-promoting state. We also summarize advanced computational tools including deep learning for data integration, spatial deconvolution, and foundation models that can infer metabolic activity from single-cell data and reconstruct cell movement over time. Results: These approaches enable researchers to identify key metabolic drivers of immune evasion and predict which checkpoints are most actionable. Conclusions: Artificial-intelligence-driven multi-omics transforms hepatocellular carcinoma research from descriptive catalogues into predictive, mechanism-based models, offering a roadmap for designing next-generation immunotherapies.