Nurix Debuts Foundation Model to Predict Novel Binders for Undruggable Targets

Nurix Therapeutics presented new data at the 2025 AACR Annual Meeting supporting the capabilities of its DEL-AI platform—a machine learning–powered system designed to accelerate discovery of small molecule binders using DNA-encoded library (DEL) data. According to Nurix, its DEL Foundation Model can predict novel binders to a wide range of therapeutically relevant targets, including proteins historically considered undruggable.

Nurix's DEL-AI Platform

The foundation model was trained on the company’s internal DEL datasets allegedly encompassing over five billion compounds screened across hundreds of protein targets and E3 ligases. By integrating this large experimental dataset with primary protein sequence data, the model learns generalizable structure–activity relationships, enabling prospective prediction of chemical binders from protein sequence alone. Notably, it's reported that the model demonstrated successful predictions even when input sequences shared only 50% similarity with proteins in the training set, and showed potential to infer binders from chemical space not represented in the original training library.

Nurix reported that virtual screening outputs closely aligned with experimental results across prospective targets. This capability, the company says, allows for proteome-wide in silico screening and faster early-stage discovery, both for internal R&D and partner-driven programs. The DEL-AI engine supports multiple workflows including degrader design, DAC (degrader-antibody conjugate) development, and synthesis automation.

The DEL Foundation Model was developed in partnership with software firm Loka and deployed on Amazon Web Services infrastructure using AWS SageMaker and MLflow.

Nurix’s broader clinical and discovery pipeline spans protein degraders, molecular glues, and immuno-oncology candidates. Partnered programs with Gilead, Pfizer, and Sanofi are in preclinical stages, while Nurix retains development rights to select programs.

Topics: AI & Digital