Autoregressive fragment-based diffusion for pocket-aware ligand design

TL;DR

This paper presents AutoFragDiff, a novel autoregressive diffusion model that generates 3D molecules conditioned on protein pockets, improving local geometry and binding affinity for ligand design.

Contribution

It introduces a fragment-based autoregressive diffusion approach with geometric vector perceptrons for pocket-aware ligand generation, enabling scaffold extension.

Findings

Enhanced local geometry of generated molecules.

Maintains high predicted binding affinity.

Capable of scaffold extension from user input.

Abstract

In this work, we introduce AutoFragDiff, a fragment-based autoregressive diffusion model for generating 3D molecular structures conditioned on target protein structures. We employ geometric vector perceptrons to predict atom types and spatial coordinates of new molecular fragments conditioned on molecular scaffolds and protein pockets. Our approach improves the local geometry of the resulting 3D molecules while maintaining high predicted binding affinity to protein targets. The model can also perform scaffold extension from user-provided starting molecular scaffold.