Prompt-based 3D Molecular Diffusion Models for Structure-based Drug Design

Ling Yang, Zhilin Huang, Xiangxin Zhou, Minkai Xu, Wentao Zhang, Yu Wang, Xiawu Zheng, Wenming Yang, Ron O. Dror, Shenda Hong, Bin CUI

Primary Area: generative models

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: Diffusion Model, Structure-based Drug Design, Molecule Generation

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Abstract: Generating ligand molecules that bind to specific protein targets via generative models holds substantial promise for advancing structure-based drug design. Existing methods generate molecules from scratch without reference or template ligands, which poses challenges in model optimization and may yield suboptimal outcomes. To address this problem, we propose an innovative prompt-based 3D molecular diffusion model named PromptDiff to facilitate target-aware molecule generation. PromptDiff leverages a curated set of ligand prompts, i.e., those with desired properties such as high binding affinity and synthesizability, to steer the diffusion model towards synthesizing ligands that satisfy design criteria. Specifically, we design a geometric protein-molecule interaction network (PMINet), and pretrain it with binding affinity signals to: (i) retrieve target-aware ligand molecules with high binding affinity to serve as prompts, and (ii) incorporate essential protein-ligand binding structures for steering molecular diffusion generation with two effective prompting mechanisms, i.e., exemplar prompting and self prompting. Empirical studies on CrossDocked2020 dataset show PromptDiff can generate molecules with more realistic 3D structures and achieve state-of-the-art binding affinities towards the protein targets, while maintaining proper molecular properties.

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: 1345