Recursion Pharma launches open‑source Nesso-1 AI tool for binding‑affinity prediction

Recursion Pharma has released Nesso-1, an open‑source (Apache‑2.0) AI model for predicting protein‑ligand binding affinity. The code and pretrained weights are publicly available on GitHub. According to its technical report, Nesso‑1 runs faster than competing models such as Boltz‑2 and shows higher accuracy on compounds that are dissimilar from its training set.

The tool can be operated via a command‑line interface and has been demonstrated on a public Tyk2 kinase dataset. Nesso‑1 joins a growing suite of AI‑driven drug‑discovery platforms that include Folding AI systems like AlphaFold, OpenFold and Boltz‑2, all aimed at accelerating the design of small‑molecule therapeutics. Small molecules remain the backbone of modern medicines, accounting for over 90% of marketed drugs, and advances in AI prediction are expected to streamline early‑stage discovery.

The release aims to provide researchers with a faster, more accessible method for evaluating binding affinity, potentially reducing the time and cost required to move from bench‑scale hits to preclinical candidates.