​​Integrating machine learning and spatial transcriptomics uncovers shared immunomodulatory deubiquitinases in MAFLD and HCC.

[UNLABELLED] Hepatocellular carcinoma (HCC) represents a malignancy with high global mortality. Metabolic dysfunction-associated fatty liver disease (MAFLD) serves as a significant contributory pathogenic factor. Deubiquitinases (DUBs), which regulate protein homeostasis, are implicated in disease progression. This study focused on identifying shared mechanisms between MAFLD and HCC, screening for key DUB genes, and constructing a novel prognostic scoring system termed the Deubiquitination Score (DUBS). The DUBS significantly stratified patient survival, with a high-DUBS indicating poor prognosis and malignant tumor progression. Patients with a low-DUBS demonstrated enhanced responses to immunotherapy and prolonged survival. Their tumors exhibited characteristics of “hot tumors,” featuring abundant immune cell infiltration and an active tumor microenvironment, accompanied by higher microsatellite instability(MSI). The core gene identified, EIF3F, exhibited superior cross-disease diagnostic value between HCC and MAFLD. In vitro and in vivo experiments demonstrated that knockdown of EIF3F inhibited the proliferation and migration of HCC cell. Furthermore, the DUBS revealed associations with sensitivity to chemotherapeutic agents. In summary, this study provided an important molecular tool and mechanistic foundation for the early screening, prognostic assessment, differential diagnosis, and prediction of immunotherapy response in HCC. It also highlighted potential directions for targeted intervention and the development of combination therapeutic strategies.

[SUPPLEMENTARY INFORMATION] The online version contains supplementary material available at 10.1007/s00439-026-02833-6.