MULTI-OBJECTIVE SYNTHESIS PLANNING BY MEANS OF MONTE CARLO TREE SEARCH

Helen Lai,Christos Kannas,Alan Kai Hassen,Emma Rydholm,Annie Westerlund,Djork-Arné Clevert,Mike Preuss,Samuel Genheden

DOI: https://doi.org/10.26434/chemrxiv-2024-3tzs5-v2

2024-10-08

Abstract:We introduce a multi-objective search algorithm for retrosynthesis planning, based on a Monte Carlo Tree search formalism. The multi-objective search allow for combining diverse set of objectives without considering their scale or weighting factors. To benchmark this novel algorithm, we employ four objectives in a total of eight retrosynthesis experiments on a PaRoutes benchmark set. The objectives ranges from simple ones based on starting material and step count to complex ones based on synthesis complexity and route similarity. We show that with the careful employment of complex objectives, the multi-objective algorithm outperforms the single-objective search and provides a more diverse set of solution that is closer to the desired objective. Our algorithm thus provides a framework for incorporating novel objectives for specific applications in synthesis planning.

Chemistry

What problem does this paper attempt to address?

The paper aims to address the problem of multi-objective synthesis planning, particularly in the context of retrosynthesis in organic chemistry. Specifically, the authors propose a multi-objective search algorithm based on Monte Carlo Tree Search (MCTS) to overcome the limitations of traditional single-objective search methods when dealing with complex objectives.

The main issues include:
1. **Limitations of existing methods**: Traditional single-objective search methods typically require multiple objectives to be linearly combined into a single objective function, and manual adjustment of weights is needed. This approach can easily introduce bias and is difficult to handle conflicting objectives.
2. **Multi-objective optimization**: To better handle diverse synthesis objectives, such as the availability of starting materials, the number of reaction steps, synthesis complexity, and path similarity, the authors propose a new multi-objective search algorithm.
3. **Experimental validation**: Through eight experiments on the PaRoutes benchmark set, the proposed multi-objective algorithm was validated to provide a broader and more diverse set of solutions compared to single-objective search methods.

In summary, the paper addresses the issue of effectively integrating multiple objectives of different scales and natures in retrosynthesis planning by introducing an improved multi-objective MCTS algorithm and demonstrates its superior performance in practical applications.