Accelerating the Generation of Molecular Conformations with Progressive Distillation of Equivariant Latent Diffusion Models

TL;DR

This paper introduces a novel accelerated sampling method for 3D molecular conformations using equivariant latent diffusion models, achieving significant speedups with minimal quality loss, enabling high-throughput molecular screening.

Contribution

The paper proposes Equivariant Latent Progressive Distillation, a new fast sampling algorithm that maintains geometric equivariance and improves sampling speed for molecular conformations.

Findings

Up to 7.5x faster sampling speed achieved.

Limited degradation in molecular stability during acceleration.

Potential for high-throughput molecular screening applications.

Abstract

Recent advances in fast sampling methods for diffusion models have demonstrated significant potential to accelerate generation on image modalities. We apply these methods to 3-dimensional molecular conformations by building on the recently introduced GeoLDM equivariant latent diffusion model (Xu et al., 2023). We evaluate trade-offs between speed gains and quality loss, as measured by molecular conformation structural stability. We introduce Equivariant Latent Progressive Distillation, a fast sampling algorithm that preserves geometric equivariance and accelerates generation from latent diffusion models. Our experiments demonstrate up to 7.5x gains in sampling speed with limited degradation in molecular stability. These results suggest this accelerated sampling method has strong potential for high-throughput in silico molecular conformations screening in computational biochemistry, drug discovery, and life sciences applications.