Scalable Non-Equivariant 3D Molecule Generation via Rotational Alignment

TL;DR

This paper introduces a scalable method for 3D molecule generation that relaxes equivariance constraints by learning sample-dependent rotations, resulting in improved efficiency and comparable quality to state-of-the-art equivariant models.

Contribution

It proposes a novel approach that aligns molecules using learned rotations, enabling non-equivariant diffusion models to generate 3D molecules efficiently.

Findings

Outperforms previous non-equivariant models in sample quality

Achieves comparable results to state-of-the-art equivariant diffusion models

Improves training and sampling efficiency

Abstract

Equivariant diffusion models have achieved impressive performance in 3D molecule generation. These models incorporate Euclidean symmetries of 3D molecules by utilizing an SE(3)-equivariant denoising network. However, specialized equivariant architectures limit the scalability and efficiency of diffusion models. In this paper, we propose an approach that relaxes such equivariance constraints. Specifically, our approach learns a sample-dependent SO(3) transformation for each molecule to construct an aligned latent space. A non-equivariant diffusion model is then trained over the aligned representations. Experimental results demonstrate that our approach performs significantly better than previously reported non-equivariant models. It yields sample quality comparable to state-of-the-art equivariant diffusion models and offers improved training and sampling efficiency. Our code is available at https://github.com/skeletondyh/RADM