Constrained synthesis planning with disconnection-aware transformer and multi-objective search

24 May 2024, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

Designing synthesis routes with shared intermediates for a set of target compounds is a common task in drug discovery. Multistep retrosynthesis tools such as AiZynthFinder are frequently used by chemists to generate possible routes. Although these tools can find solved routes for a majority of target compounds, they may not generate routes which comply with specific bond constraints. Such bond constraints could be defined in the project to obtain synthesis routes with common intermediates for the set of compounds. Here, we present a novel retrosynthesis approach which aims to generate routes in the feasible region defined by these constraints. The constraints are divided into bonds to break and bonds to freeze. First, we introduce a filter in the search which discards any reaction that violates the frozen bond constraints. Second, we benchmark four possible strategies for breaking selected bonds in the search for synthetic routes. We show that a combination of disconnection-aware transformer and multi-objective search generates routes which satisfy bond constraints for more targets in the PaRoutes dataset compared to the standard search (75.57% vs. 54.80%). Finally, we apply the strategy on a set of drug compounds to exemplify a real-world scenario. Using our novel approach, we can build a short joint synthesis route which satisfies the given bond constraints and covers eight of the ten compounds.

Keywords

retrosynthesis
monte carlo tree search
MCTS
chemformer
transformer
artificial intelligence
machine learning
template-based

Supplementary materials

Title
Description
Actions
Title
Supplementary material
Description
Figures, tables and text supporting the main text
Actions

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.