TorSeq: Torsion Sequential Modeling for Molecular 3D Conformation Generation
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: conformer generation, sequential model, geometric deep learning
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: Molecular 3D conformer generation via sequential modeling over torsion angles.
Abstract: In the realms of chemistry and drug discovery, the generation of 3D low-energy molecular conformers is critical. While various methods, including deep generative and diffusion-based techniques, have been developed to predict 3D atomic coordinates and molecular geometry elements like bond lengths, angles, and torsion angles, they often neglect the intrinsic correlations among these elements. This oversight, especially regarding torsion angles, can produce less-than-optimal 3D conformers in the context of energy efficiency. Addressing this gap, we introduce a method that explicitly models the dependencies of geometry elements through sequential probability factorization, with a particular focus on optimizing torsion angle correlations. Experimental evaluations on benchmark datasets for molecule conformer generation underscore our approach's superior efficiency and efficacy.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 2676
Loading