Cellarity published an AI-driven toxicogenomics platform, ToxPredictor, supported by a DILImap transcriptomics library profiling 300 compounds, to predict dose-related drug-induced liver injury (DILI). The model reportedly achieved 88% sensitivity at 100% specificity in validation and identified Phase III safety failures missed by animal studies. Cellarity made the resource available for non-commercial use and described its method in Nature Communications. If broadly replicable, the approach could shift preclinical safety assessment by integrating human hepatocyte transcriptional signatures into early go/no-go decisions. Biotech and pharma safety teams will test whether transcriptomics-guided predictions can reduce late-stage attrition driven by hepatotoxicity.