DirectMultiStep: Direct Route Generation for Multistep Retrosynthesis

TL;DR

This paper introduces transformer-based models for direct multistep retrosynthesis route generation, significantly improving accuracy and efficiency over traditional iterative methods, and demonstrating strong generalization to unseen drugs.

Contribution

Proposes a novel transformer-based approach that directly generates multistep synthetic routes, outperforming state-of-the-art methods and enabling incorporation of additional constraints for better predictions.

Findings

Outperforms state-of-the-art on PaRoutes dataset with 1.9x and 3.1x Top-1 accuracy improvements.

Models generalize well to FDA-approved drugs not in training data.

Incorporating constraints improves accuracy and reduces model size.

Abstract

Traditional computer-aided synthesis planning (CASP) methods rely on iterative single-step predictions, leading to exponential search space growth that limits efficiency and scalability. We introduce a series of transformer-based models, that leverage a mixture of experts approach to directly generate multistep synthetic routes as a single string, conditionally predicting each transformation based on all preceding ones. Our DMS Explorer XL model, which requires only target compounds as input, outperforms state-of-the-art methods on the PaRoutes dataset with 1.9x and 3.1x improvements in Top-1 accuracy on the n$_1$ and n$_5$ test sets, respectively. Providing additional information, such as the desired number of steps and starting materials, enables both a reduction in model size and an increase in accuracy, highlighting the benefits of incorporating more constraints into the prediction process. The top-performing DMS-Flex (Duo) model scores 25-50% higher on Top-1 and Top-10 accuracies for both n$_1$ and n$_5$ sets. Additionally, our models successfully predict routes for FDA-approved drugs not included in the training data, demonstrating strong generalization capabilities. While the limited diversity of the training set may affect performance on less common reaction types, our multistep-first approach presents a promising direction towards fully automated retrosynthetic planning.