Using AI to combine many NAMs with drug exposure and metabolism to model DILI for modern molecules.

Drug-induced liver injury (DILI) remains a leading cause of clinical attrition, a challenge exacerbated by reliance on historical benchmarks unrepresentative of modern chemotypes. To address this, we developed an AI-driven DILI prediction platform that integrates diverse New Approach Methodologies including high-content imaging of 2D primary human hepatocytes and 2D+ Multi-cell Hepatic Systems with predicted drug metabolism and exposur. This approach is underpinned by a proprietary dataset exceeding 130,000 molecules and a meticulously curated clinical benchmark of hundreds of modern drugs, utilizing LLM-assisted curation to ensure accurate human exposure and incidence data. We demonstrate that iteratively integrating these diverse NAMs with drug exposure significantly enhances predictive performance on modern molecules compared to models relying solely on dose and simple viability assays. This presentation will highlight how the integration of exposure and metabolism data is critical for accurate DILI risk assessment and showcase case studies where this platform successfully differentiated between modern clinically safe and toxic analogs.