Retrosynthesis Prediction via Search in (Hyper) Graph

Zixun Lan, Binjie Hong, Jiajun Zhu, Zhenfu Liu, Zuo Zeng, Fei Ma

Keywords: Retrosynthesis Prediction via \\ Search in (Hyper) Graph

Abstract: Retrosynthesis prediction is a fundamental challenge in organic synthesis, involving the prediction of reactants based on a given core product. Recently, semi-template-based methods and graph-edits-based methods have achieved good performance in interpretability and accuracy. However, their mechanisms still fail to predict complex reactions, e.g., reactions with multiple reaction center or attaching the same leaving group to more than one atom. Hence, we propose, a semi-template-based method, the \textbf{Retro}synthesis via \textbf{S}earch \textbf{i}n (Hyper) \textbf{G}raph (RetroSiG) framework to alleviate these limitations. In this paper, we cast the reaction center identification and the leaving group completion as search in the product molecular graph and leaving group hypergraph respectively. RetroSiG has several advantages as a semi-template-based method: First, RetroSiG is able to handle the complex reactions mentioned above with its novel search mechanism. Second, RetroSiG naturally exploits the hypergraph to model the implicit dependencies between leaving groups. Third, RetroSiG makes full use of the prior, i.e., one-hop constraint. It reduces the search space and enhances overall performance. Comprehensive experiments demonstrate that RetroSiG achieves a competitive result. Furthermore, we conduct experiments to show the capability of RetroSiG in predicting complex reactions. Ablation experiments verify the effectiveness of individual components, including the one-hop constraint and the leaving group hypergraph.

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.

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

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: 2745