Do Deep Learning Methods Really Perform Better in Molecular Conformation Generation?

TL;DR

This paper challenges the assumption that deep learning methods outperform traditional approaches in molecular conformation generation by demonstrating a simple, parameter-free clustering algorithm that matches or exceeds their performance on standard benchmarks.

Contribution

The authors introduce a simple clustering-based algorithm that rivals deep learning methods in molecular conformation generation, questioning the claimed superiority of deep learning approaches.

Findings

The clustering algorithm performs comparably or better than deep learning methods on GEOM-QM9 and GEOM-Drugs benchmarks.

Traditional, simple methods can be competitive with complex deep learning models in MCG.

The study encourages revisiting and revising current deep learning approaches in molecular conformation generation.

Abstract

Molecular conformation generation (MCG) is a fundamental and important problem in drug discovery. Many traditional methods have been developed to solve the MCG problem, such as systematic searching, model-building, random searching, distance geometry, molecular dynamics, Monte Carlo methods, etc. However, they have some limitations depending on the molecular structures. Recently, there are plenty of deep learning based MCG methods, which claim they largely outperform the traditional methods. However, to our surprise, we design a simple and cheap algorithm (parameter-free) based on the traditional methods and find it is comparable to or even outperforms deep learning based MCG methods in the widely used GEOM-QM9 and GEOM-Drugs benchmarks. In particular, our design algorithm is simply the clustering of the RDKIT-generated conformations. We hope our findings can help the community to revise the deep learning methods for MCG. The code of the proposed algorithm could be found at https://gist.github.com/ZhouGengmo/5b565f51adafcd911c0bc115b2ef027c.