Abstract 3762: DEL-AI: Proteome-wide <i>in silico</i> screening of multi-billion compound libraries using machine learning foundation models

Abstract DNA-encoded libraries (DEL) are a transformative technology in small-molecule discovery and the engine driving Nurix’s small molecule and degrader pipelines for novel targets. DEL platforms allow efficient screening of libraries of billions of unique molecules against a diverse set of biological targets. This approach generates datasets well suited for machine learning (ML) methods trained to identify patterns and relationships, which then guides the design of novel compounds. Reported methods have focused on the development of ML models built from experimental datasets and thus require a successful DEL screen as a prerequisite. However, executing high-quality DEL screens is an extremely resource intensive process limited by physical factors such as protein production and availability. As such, the scale at which experimental screens and their derivative ML models can be applied is inherently limited, and some targets remain out of reach. Recently, we have developed an ML technology, DEL-AI, which overcomes these limitations and is able to accurately predict experimental DEL data given only a target protein sequence. Nurix’s DEL-AI Platform is a machine learning framework designed to optimize DEL hit identification and predictive power for downstream applications without the need to first screen a protein target experimentally. Trained on the high-quality and extensive DEL archive at Nurix, our platform accurately predicts binding of multi-billion molecule DEL libraries by learning the fundamental relationships between chemical structure and protein sequence. Under this paradigm, we are able to conduct large-scale virtual Lead ID for entire therapeutic target classes. Citation Format: Elena Caceres, Cristiana Carpinteiro, Manuel Ravasqueira, Mridula Bontha, Brandon Bravo, Graham Carlson, Jim Davis, Ketki Dhamnaskar, John Eichenseer, Telmo Felgueira, Jamie Furneisen, Heta Gandhi, Jennifer Greufe, Chad Hewitt, Tim Kane, Tatiana Kennedy, Buckley Kolhauff, Emily Kruger, Ajay Kulkarni, Evan Landstrom, Calvin Larson, Emily Low, Marie Malone, Ratul Mukerji, Chris B. Phelps, Patricia Rocha, Sushmitha Regulapati, Jorge Sampaio, Nick Sanchez, Jose Santos, Julie Sheung, Vaishali Sridhar, Jeffrey Wu, Lik Hang Yuen, Stephanie Yung, Kiril Zelenkovski, Paul Novick. DEL-AI: Proteome-wide in silico screening of multi-billion compound libraries using machine learning foundation models [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3762.