Abstract
Predictive safety risk biomarkers for drug-induced liver injury (DILI) are critically needed to enhance patient risk stratification and minimize adverse outcomes. This study aims to identify genomic markers associated with increased mortality in toxic liver disease, using the KEM® (Knowledge Extraction and Management) explainable Artificial Intelligence platform. From 225 participants diagnosed with toxic liver disease within the UK Biobank cohort, data were consolidated, including survival outcomes, clinical phenotypes, comorbidities, and 36 394 genomic single nucleotide polymorphisms (SNPs) focusing on liver-related pathways. Fifteen SNPs were found to be significantly associated with increased mortality risk, notably rs73158145 in the PRKAG2 gene. Predictive models built on these selected SNPs achieved a mean accuracy of 85%, outperforming models without pre-selection (68.9% accuracy). Further validation in independent cohorts is planned to confirm the clinical relevance of these biomarkers.