Integrated multi-omics and machine learning identify EFNA3 as a key biomarker of tumor invasion

Gastric cancer remains a major global health burden, and its pronounced molecular heterogeneity hampers progress in precise subtyping and targeted therapy. Invasion-related programs are considered central drivers of malignant progression, yet cross-scale and pan-cancer evidence remains limited. This study aimed to stratify gastric cancer samples by invasion and, within a multi-omics framework, comprehensively assess their clinical features to identify stable biomarkers with translational potential.

By integrating multiple machine-learning algorithms with multi-omics data, we identified EFNA3 as a core gene closely associated with the invasive phenotype. We further assessed its distribution in malignant cellular subpopulations, its biological associations, and its relevance across multiple cancer types. In vitro and in vivo experiments were performed to validate its functional roles.

EFNA3 was enriched in specific malignant cellular subpopulations and involved in regulating malignant cell development, accompanied by enhanced cell-adhesion, DNA-metabolic, and cell-cycle programs. EFNA3 was consistently overexpressed across multiple cancer types and associated with tumor progression and poor survival. Its expression also coincided with reduced infiltration of effector immune cells and downregulation of immune checkpoints, indicating an immunosuppressive microenvironment. In vitro and in vivo experiments further demonstrated that EFNA3 promotes gastric cancer cell proliferation, migration, and tumor growth.

Collectively, EFNA3 is closely linked to pro-tumor processes and immune evasion in gastric and multiple other cancers, supporting its value as a subtype biomarker and potential therapeutic target.