insitro and Eli Lilly Partner to Develop AI-Powered ADMET Prediction Models for Small Molecule Drug Discovery
核心洞察
insitro (搜索) and Eli Lilly announced a collaboration to develop advanced machine learning models that can accurately predict key pharmacological properties of small molecules, including their behavior in vivo.
The partnership will leverage Lilly's proprietary preclinical data from decades of drug discovery programs to train AI models for predicting ADMET (搜索) properties, potentially reducing development timelines and costs.
The models will be integrated into Lilly TuneLab, a new drug discovery platform designed to accelerate medicine development by providing biotechnology companies access to powerful machine learning tools.
insitro (搜索), a pioneer in machine learning for drug discovery and development, announced a new collaboration with Eli Lilly and Company to develop advanced machine learning models that can accurately predict key pharmacological properties of small molecules, including their behavior in vivo. This partnership aims to address longstanding challenges in drug development where such properties have traditionally been slow and costly to determine through experimental methods.
The collaboration seeks to overcome industry-wide challenges in small molecule design, where insufficient data or inadequate models have hindered progress in building next-generation predictive tools that save time and reduce experimentation cycles.
Leveraging Decades of Proprietary Data
Under the agreement, insitro (搜索) will build advanced machine learning models and train them on Lilly's proprietary preclinical data, leveraging a rich set of in vitro and in vivo measurements from a vast array of compounds with established ADMET (搜索) (Absorption, Distribution, Metabolism, Excretion, and Toxicity) properties. This dataset, derived from decades of Lilly's drug discovery programs, represents what the companies describe as a world-class collection in terms of quality, consistency and scale.
"The rapid design of safe and effective small molecules has long been a holy grail in drug discovery, but has been stymied by the unpredictability of key pharmacological properties, such as a molecule's behavior in vivo," said Daphne Koller, Ph.D., founder and CEO of insitro (搜索). "AI can address this challenge, but only with robust, coherent, and consistently collected data on advanced molecules, data that are very rarely found."
Addressing Cost and Time Challenges
The models being developed are designed to improve the efficiency of hit-to-lead and lead optimization efforts by predicting multiple ADMET (搜索) properties including the in vivo behavior of small molecules. Traditional approaches to optimizing pharmacokinetics can often take years and cost tens of millions of dollars, making this AI-driven approach particularly significant for the industry.
Philip Tagari, Chief Scientific Officer of insitro (搜索), emphasized the potential impact: "These models have the potential to be a game-changer by giving researchers an elegant and powerful way to zero in on drug-like chemical structures at the earliest stages. Small molecules that reach the right tissue at the optimal concentration for the right duration result in better patient outcomes."
Integration with Lilly TuneLab Platform
The machine learning models will be available to insitro (搜索) and Lilly, as well as their partners, including biotech companies that partner with Lilly TuneLab, a new drug discovery platform announced alongside this collaboration. Lilly TuneLab is designed to accelerate the development of new medicines by providing biotechs access to powerful machine learning models and is part of the Lilly Catalyze360 model, a comprehensive approach to empower early-stage biotechs.
The platform is built on a federated learning infrastructure and hosted by a third-party provider, with both Lilly's and its partners' data remaining separate and private. The models will be continuously updated as the dataset continues to expand.
Expanding Partnership
This collaboration expands the relationship between insitro (搜索) and Lilly, which was initially announced in 2024 and focused on Lilly's siRNA delivery and antibody discovery capabilities to enable insitro's emerging pipeline in metabolic diseases (搜索).
The novel ADMET (搜索) models will become a critical component of insitro (搜索)'s end-to-end AI capability for small molecule chemistry, including machine learning and physics-based in silico screening, affinity machine learning models from proprietary DNA-encoded libraries, and an active learning medicinal chemistry engine that together comprise insitro's ChemML platform.
With more than $700 million in capital raised to date, insitro (搜索) is building a "pipeline through platform" approach with a focus on metabolic disease and neuroscience, aiming to deploy AI models to run smaller, better powered trials that enroll patients who can benefit most.