The abstract is read at the publisher; the summary is JClub's.
Savant predicts protein-ligand dissociation constants from protein amino-acid sequence plus ligand SMILES, ranking ahead of all other 3-dimensional structure-free methods on an independent benchmark.
- Why it matters: Early drug discovery needs fast binding-affinity estimates, but generative chemistry and large catalogs outpace rigorous evaluation, while leading approaches require GPU-heavy co-folding.
- What they did: The authors built a leakage-resistant benchmark using only post-assembly published measurements and chemically unlike molecules, then compared Savant against other 3-dimensional structure-free methods.
- The result: Savant runs on a CPU without docking or 3D folding, matches GPU-based methods for single-protein ranking, and screens 100,000 compounds in minutes on commodity hardware.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.