NEAT-POCKET: Pocket-Conditioned Autoregressive 3D Molecular Generation with a Neighborhood-Guided Set Transformer
NEAT-POCKET is a pocket-conditioned extension of the autoregressive NEAT model for 3D molecular generation. It generates molecules atom by atom in protein pocket environments while preserving atom permutation invariance and explicitly modeling hydrogen atoms. Benchmarks on CrossDocked and SPINDR datasets show competitive structure-based generation performance and substantially faster sampling than existing baselines. It also enables pocket-conditioned fragment completion for lead optimization and scaffold elaboration.
The model extends the NEAT autoregressive framework to condition on protein pockets, maintaining permutation invariance and explicit hydrogen modeling. Its faster sampling compared to baselines suggests efficiency gains in structure-based generation. The ability to perform fragment completion indicates flexibility for partial molecule generation tasks.
This research targets early-stage drug discovery by generating ligands within binding pockets. Faster sampling and fragment completion capabilities could accelerate lead optimization workflows in pharmaceutical R&D.
The framework may reduce time and cost in early drug discovery by enabling rapid generation and optimization of candidate molecules within target pockets.
Potential next signals include adoption of NEAT-POCKET in drug discovery pipelines, further benchmarks on additional datasets, or integration with other molecular design tools.