Multi-omics and single-cell approaches reveal molecular subtypes and key cell interactions in hepatocellular carcinoma.

Introduction: Hepatocellular carcinoma is a highly aggressive and heterogeneous malignancy with limited understanding of its heterogeneity. Methods: In this study, we applied ten multi-omics classification algorithms to identify three distinct molecular subtypes of HCC (C1-C3). To further explore the immune microenvironment of these molecular subtypes, we leveraged single-cell transcriptomic data and employed CIBERSORTx to deconvolute their immune landscape. Results: Among them, C3 exhibited the worst prognosis, whereas C1 and C2 were associated with relatively better clinical outcomes. Patients in the C3 group exhibited a high burden of copy number variations, mutation load, and methylation silencing. Our results revealed that compared to C1 and C2, C3 had a lower proportion of hepatocytes but a higher proportion of cholangiocytes and macrophages. Through analyses of hepatocyte, cholangiocyte, and macrophage subpopulations, we characterized their functional states, spatial distribution preferences, evolutionary relationships, and transcriptional regulatory networks, ultimately identifying cell subpopulations significantly associated with patient survival. Furthermore, we identified key ligand-receptor interactions, such as APOA1-TREM2 and APOA2-TREM2 in hepatocyte-macrophage crosstalk, and VTN-PLAUR in cholangiocyte-macrophage communication. Discussion: Finally, we employed machine learning methods to construct a prognostic model for HCC patients and identified novel potential compounds for high risk patients. In summary, our novel multi-omics classification of HCC provides valuable insights into tumor heterogeneity and prognosis, offering potential clinical applications for precision oncology.