Cellarity published a toxicogenomics resource and an AI model, ToxPredictor, designed to predict drug‑induced liver injury (DILI) using transcriptomic signatures from primary human hepatocytes. The company reported 88% sensitivity at 100% specificity in validation, and said the tool flagged several Phase III safety failures that animal models missed. The DILImap resource covers transcriptional signatures for roughly 300 DILI‑linked compounds across doses. Cellarity’s open‑source dataset and model aim to help discovery teams make earlier go/no‑go decisions on hepatic risk and to reduce late‑stage attrition. Biopharma safety groups and CROs will evaluate real‑world applicability across modalities and exposures; the model’s dose‑specific predictions and mechanistic readouts could influence preclinical safety pipelines and regulatory toxicology discussions.