Cellarity published validation of ToxPredictor, an AI toxicogenomics model built on DILImap—transcriptomic signatures from primary human hepatocytes exposed to 300 DILI‑linked compounds. The firm reported 88% sensitivity at 100% specificity in retrospective validation and said the model identified Phase III safety failures missed by animal studies, positioning the tool as a human‑centric screen to de‑risk late‑stage liver toxicity. Cellarity opened parts of the resource for non‑commercial use and framed ToxPredictor as a way to detect dose‑related liver injury earlier in drug discovery. The company, which is also advancing a first clinical candidate (CLY‑124), said integrating transcriptomics with active‑learning AI will improve early safety decisions and reduce downstream clinical attrition. Clarification: Drug‑induced liver injury (DILI) is a major cause of clinical trial failure and post‑marketing withdrawals, and predictive human models aim to reduce late‑stage safety surprises.