Boltz‑2 and physics‑based models: AI advances binding‑affinity prediction

Researchers from MIT and Recursion released Boltz‑2, an open‑source model that improved speed and accuracy for predicting binding affinities—delivering results in seconds and ranking top at CASP16 benchmarks. Industry voices contrast large quantitative models (LQMs), grounded in physics, with text‑based LLMs for drug discovery. The combination of structural co‑folding tools (AlphaFold3, RoseTTAFold) and fast affinity predictors aims to cut experimental screening and accelerate hit‑to‑lead timelines.