Machine learning-driven multi-omics integration uncovers a senescence associated molecular axis in HCC

Hepatocellular carcinoma (HCC) exhibits profound molecular heterogeneity and aberrant cellular senescence. This study systematically dissects the senescence-associated molecular landscape to identify key regulators driving HCC progression and immune evasion.

Integrating multi-cohort transcriptomic datasets, we developed a robust prognostic signature using 101 machine-learning models, identifying prognostic signature. We employed preliminary proteomic, exploratory metabolomic, and single-cell RNA sequencing (scRNA-seq) analyses to explore multi-omics alterations. The functional senescence status and MCM7 were validated in a clinical HCC cohort by RT-qPCR, Western blotting, immunohistochemistry, and multiplex immunofluorescence (mIF). Causality was established using in vitro functional assays in HepG2 cells.

A 12-gene random survival forest (RSF) signature accurately predicted patient survival across independent cohorts. MCM7 emerged as a central senescence-associated driver. ScRNA-seq and mIF confirmed MCM7 characterizes a highly proliferative, clonally expanding subset of CD8⁺ T cells within the tumor microenvironment. In vitro, MCM7 knockdown significantly inhibited HepG2 cell proliferation and upregulated senescence enforcers p16 and p21, whereas overexpression facilitated evasion. Additionally, TIDE analysis revealed that high-risk patients exhibited elevated immune evasion potential, predicting poor immunotherapy response.

This integrative multi-omics framework uncovers an MCM7 MCM7-driven senescence-associated axis promising HCC progression and immune dysfunction, offering a robust tool for prognostic stratification and novel therapeutic insights.