Genesis reported that its Pearl system beat six cofolding models on an OpenBind protease benchmark.
Pearl was tested on the OpenBind EV-A71 2A protease benchmark of 802 follow-on ligand-protein complexes, against AlphaFold3, Boltz-1, Boltz-2, OpenFold3-p2, Protenix, RosettaFold 3 and docking tools. Zero-shot success was 78%, pocket-conditioned 85% and redocking 89%. At sub-angstrom accuracy Pearl reached 60% zero-shot.
Genesis Therapeutics (Genesis Molecular AI) is an AI small-molecule discovery company whose GEMS platform and Pearl model support partnerships with Gilead and Incyte.
This sits in the Design & Stress-Test stage of the AI use-case lifecycle: Named examples from the tools pharma teams use to scope, prototype and benchmark an AI use case before committing to build it — data-catalog, sandbox and evaluation platforms.
Published by Genesis Therapeutics (Genesis Molecular AI) on its own site — read the original case study on genesis.ml ↗
Related use cases
- 01Gilead and Genesis deployed the GEMS platform to find small molecules against targets that lack training data.
- 02Incyte and Genesis expanded their collaboration to at least five new targets, with Incyte data used to train GEMS.
- 03Genesis introduced Pearl, a generative 3D foundation model for protein-ligand structure prediction.
- 04Novo Nordisk Transforms Clinical Trials With AI
- 05How Regeneron is discovering new treatments with AI
- 06How AI is changing drug discovery