Consensus artificial intelligence-driven prognostic signature for predicting the prognosis of hepatocellular...

Hepatocellular carcinoma (HCC), a leading cause of global cancer mortality, requires molecular stratification to advance precision oncology. This study developed a consensus artificial intelligence-derived prognostic signature (CAIPS) by integrating ten machine learning algorithms (101 methods) across six multi-center HCC cohorts (n = 1110). The optimized seven-gene CAIPS, constructed using StepCox[both] and GBM, demonstrated superior prognostic accuracy over traditional clinical parameters and 150 published signatures. Multi-omics profiling linked high CAIPS scores to metabolic pathway dysregulation and genomic instability, whereas low CAIPS scores predicted enhanced therapeutic responsiveness to transcatheter arterial chemoembolization, targeted therapies, and immunotherapy. Screening of CTPR, PRISM, and Connectivity Map databases prioritized Irinotecan and BI-2536 as candidate therapeutics for high-CAIPS patients. Functional validation revealed that PITX1 knockdown significantly suppressed HCC cell proliferation, invasion, migration, and xenograft tumor growth, mechanistically attributed to Wnt/β-catenin signaling inhibition. In vitro experiments revealed that Irinotecan and BI-2536 exhibit high potential as anti-HCC drugs. Collectively, CAIPS serves as a robust multi-dimensional biomarker system for risk stratification, therapy optimization, and personalized HCC management. The concurrent identification of Irinotecan and BI-2536 as targeted agents and PITX1-mediated pathway regulation establishes actionable frameworks for precision oncology.