3D Molecule Generation from Rigid Motifs via SE(3) Flows

TL;DR

This paper introduces a novel method for 3D molecule generation using rigid motifs and SE(3)-equivariant models, achieving faster, more stable, and more compact molecular representations compared to atom-based approaches.

Contribution

It extends molecular generation to rigid-body motifs in 3D using SE(3) flows, improving efficiency and stability over existing atom-based methods.

Findings

Achieves comparable or better results than state-of-the-art methods.

Surpasses in atom stability on GEOM-Drugs benchmark.

Reduces generation steps by 2x to 10x and compresses molecular representations 3.5x.

Abstract

Three-dimensional molecular structure generation is typically performed at the level of individual atoms, yet molecular graph generation techniques often consider fragments as their structural units. Building on the advances in frame-based protein structure generation, we extend these fragmentation ideas to 3D, treating general molecules as sets of rigid-body motifs. Utilising this representation, we employ SE(3)-equivariant generative modelling for de novo 3D molecule generation from rigid motifs. In our evaluations, we observe comparable or superior results to state-of-the-art across benchmarks, surpassing it in atom stability on GEOM-Drugs, while yielding a 2x to 10x reduction in generation steps and offering 3.5x compression in molecular representations compared to the standard atom-based methods.