CoarsenConf: Equivariant Coarsening with Aggregated Attention for Molecular Conformer Generation

• DOI: 10.1021/acs.jcim.4c01001
• Publication Date: 2024-12-17T15:18:58Z
• Authors: Danny Reidenbach; Aditi S. Krishnap...
• Keywords: Conformational isomerism; Cheminformatics; Equivariant map; Generative model; Representation

Molecular conformer generation (MCG) is an important task in cheminformatics and drug discovery. The ability to efficiently generate low-energy 3D structures can avoid expensive quantum mechanical simulations, leading accelerated virtual screenings enhanced structural exploration. Several generative models have been developed for MCG, but many struggle consistently produce high-quality conformers meaningful downstream applications. To address these issues, we introduce CoarsenConf, which coarse-grains molecular graphs based on torsional angles integrates them into SE(3)-equivariant hierarchical variational autoencoder. Through equivariant coarse-graining, aggregate the fine-grained atomic coordinates of subgraphs connected via rotatable bonds, creating a variable-length coarse-grained latent representation. Our model uses novel aggregated attention mechanism restore from representation, enabling efficient accurate conformers. Furthermore, evaluate chemical biochemical quality our generated multiple applications, including property prediction large-scale oracle-based protein docking. Overall, CoarsenConf generates more ensembles compared prior models.