From Holo Pockets to Electron Density: GPT-style Drug Design with Density

Researchers introduced EDMolGPT, a decoder-only autoregressive model that generates molecules from low-resolution electron density point clouds.

The model was tested across 101 biological targets and can use density from computational or experimental sources such as cryo-EM and X-ray.

By conditioning on physically grounded density instead of empty binding pockets, EDMolGPT better captures site flexibility and supports de novo drug design.