Multi-omics-driven Molecular Subtyping of Hepatocellular Carcinoma and Comprehensive Functional Profiling of Post-subtyping Marker Genes

Abstract

Hepatocellular carcinoma (HCC) is characterized by profound molecular heterogeneity and dismal prognosis; however, current classification systems relying on single-omics data lack consensus and fail to guide personalized therapy robustly. To bridge this gap, we aimed to establish a stable, multi-omics subtyping framework that captures the full spectrum of HCC biological diversity. By integrating transcriptomic (mRNA, lncRNA, miRNA), DNA methylation, mutation, and clinical profiles from TCGA, an ensemble strategy combining 10 distinct clustering algorithms was employed to identify robust HCC subtypes. Subtype-specific prognostic models were constructed via Cox regression, and molecular signatures were validated using single-cell RNA sequencing (scRNA-seq) and immunotherapy response algorithms (TIDE). We identified two distinct molecular subtypes (CS1 and CS2) with significant survival disparities. The CS2 subtype exhibited an immune “hot” phenotype, characterized by high immune checkpoint expression and NK cell infiltration, correlating with superior predicted response to immune checkpoint blockade (ICB); conversely, CS1 presented an immune “cold” phenotype with distinct metabolic alterations. Single-cell analysis revealed cell-type-specific expression patterns of marker genes, and drug sensitivity profiling identified subtype-specific vulnerabilities to conventional chemotherapeutics. By associating molecular subtypes with distinct phenotypic features, our study helps bridge the gap between multi-omics insights and precision HCC management. Our multi-omics framework provides a robust classification system that not only deciphers the molecular heterogeneity of HCC but also stratifies patients based on their predicted immune and therapeutic profiles, offering a roadmap for precision oncology.