AI-Powered Drug Discovery
AlmaDyne - AI that maps protein dynamics to uncover new drug targets.
Scroll
Our mission is to unlock undruggable targets by mapping the full conformational landscape of disease-relevant proteins in their native biological environment.
Proteins are not static — they breathe, flex, and shift between conformations. Traditional methods photograph only one frozen state, missing the transient pockets where drugs can bind.
AlmaDyne models the full dynamic range of a protein in its native membrane or cytoplasmic environment — not just a crystal structure snapshot.
Transient conformations expose binding sites invisible to conventional structure-based design — opening entirely new druggable surfaces.
AI-designed candidates are tested in vitro and in vivo with Columbia and King's College London, closing the loop back into the model.
Without any published structural data, AlmaDyne identified active binders from a small compound set — two chemically distinct scaffolds confirmed, with direct target engagement validated by radioligand binding assay.
40× higher hit rate vs industry benchmarks — demonstrated in the CACHE challenge, published in Nature.
Two chemically distinct scaffolds confirmed, validating the platform's ability to generate diverse chemotypes from first principles.
Direct target engagement confirmed by competitive radioligand binding assay using ³H-SNAP94847.
MCHR1 receptor with Alma-1 candidate (yellow) docked in the binding pocket identified by AlmaDyne — without access to any published structural data.