insitro, a leader in applying machine learning to drug discovery and development, has announced a strategic collaboration with Eli Lilly and Company to create advanced machine learning models that can predict critical pharmacological properties of small molecules, particularly their behavior in living organisms. This partnership aims to overcome long-standing challenges in drug development, where determining such properties through traditional lab experiments is time-consuming and expensive. The collaboration leverages insitro’s deep expertise in computational biology and artificial intelligence, combined with Lilly’s extensive, high-quality preclinical data. This dataset includes decades of in vitro and in vivo measurements from a vast number of compounds with well-characterized ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) profiles—making it one of the most comprehensive and consistent collections of its kind. The goal is to build next-generation predictive models capable of accurately forecasting key pharmacokinetic and safety characteristics early in the drug discovery process. By doing so, the models will help researchers design small molecules with optimal drug-like properties from the outset, significantly reducing the need for costly and time-intensive animal studies and accelerating the path from concept to clinical candidate. Daphne Koller, Ph.D., founder and CEO of insitro, highlighted the significance of the partnership, noting that predicting how small molecules behave in the body has long been a major bottleneck in drug development. She emphasized that AI can only deliver transformative results when trained on robust, well-curated data—something Lilly’s dataset uniquely provides. “We’re excited to bring our machine learning capabilities to Lilly’s world-class data to build best-in-class models that can deliver better drugs faster,” Koller said. Philip Tagari, Chief Scientific Officer at insitro, added that the new models could revolutionize early-stage drug design by enabling researchers to identify promising chemical structures with greater precision and confidence. “These models will help us zero in on drug-like molecules early, improving the likelihood of success and ultimately leading to better patient outcomes,” he said. The machine learning models will be used internally by both insitro and Lilly, as well as by biotech partners accessing Lilly TuneLab™, a new platform launched as part of Lilly’s Catalyze360 initiative. Designed to support early-stage biotech innovation, Lilly TuneLab uses a federated learning framework to securely analyze data across organizations while keeping proprietary information private and separate. The models will be continuously refined as more data is added, enhancing their accuracy and utility over time. They represent a core component of insitro’s broader ChemML platform, which integrates AI-driven screening, physics-based simulations, affinity prediction from DNA-encoded libraries, and an active learning engine for medicinal chemistry. This collaboration builds on a prior partnership announced in 2024 focused on siRNA delivery and antibody discovery, expanding insitro’s efforts in metabolic diseases. With over $700 million in funding, insitro is advancing an end-to-end AI-powered drug discovery pipeline, aiming to bring more effective therapies to patients faster.