Isomorphic Labs introduces the Iso DDE, a unified computational drug-design system progressing beyond AlphaFold 3 in predictive accuracy. The engine more than doubles AlphaFold 3's accuracy on the "Runs N' Poses" generalization benchmark for protein-ligand structures, predicts binding affinities exceeding physics-based methods at a fraction of time and cost, and identifies novel binding pockets from amino acid sequence alone.

Since AlphaFold 3's 2024 release with Google DeepMind, the field has advanced rapidly. While AlphaFold 3 transformed protein-ligand prediction and its database serves over 3 million researchers across 190 countries, it struggled to generalize to unexplored biomolecular regions. Iso DDE demonstrates a step change, successfully modeling induced fits and cryptic pocket opening even for systems distant from training data.

For complex biologics, Iso DDE outperforms AlphaFold 3 by 2.3x and Boltz-2 by 19.8x in high-fidelity antibody-antigen prediction, effectively unlocking de novo antibody design. It surpasses all deep-learning methods and physics-based approaches like FEP on binding affinity benchmarks, enabling rapid molecule optimization.

Iso DDE identifies ligandable pockets in seconds, approaching experimental techniques like fragment-soaking. This revealed new druggable pockets on cereblon, previously believed to have only one principal binding site for 15 years.