Molecular Conformation
Molecular conformation, the three-dimensional arrangement of atoms in a molecule, is crucial for understanding its properties and functions. Current research focuses on developing accurate and efficient methods for generating and analyzing molecular conformations, employing diverse approaches such as diffusion models, graph neural networks, and flow-based generative models, often incorporating physical constraints and symmetries (e.g., SE(3) invariance). These advancements are significantly impacting fields like drug discovery and materials science by enabling faster and more accurate virtual screening, protein design, and reaction pathway prediction. The development of robust benchmarks and datasets is also a key area of focus, facilitating the comparison and improvement of different methods.
Papers
ET-Flow: Equivariant Flow-Matching for Molecular Conformer Generation
Majdi Hassan, Nikhil Shenoy, Jungyoon Lee, Hannes Stark, Stephan Thaler, Dominique Beaini
Pushing the Limits of All-Atom Geometric Graph Neural Networks: Pre-Training, Scaling and Zero-Shot Transfer
Zihan Pengmei, Zhengyuan Shen, Zichen Wang, Marcus Collins, Huzefa Rangwala