Energy-based View of Retrosynthesis
read the original abstract
Retrosynthesis -- the process of identifying a set of reactants to synthesize a target molecule -- is of vital importance to material design and drug discovery. Existing machine learning approaches based on language models and graph neural networks have achieved encouraging results. In this paper, we propose a framework that unifies sequence- and graph-based methods as energy-based models (EBMs) with different energy functions. This unified perspective provides critical insights about EBM variants through a comprehensive assessment of performance. Additionally, we present a novel dual variant within the framework that performs consistent training over Bayesian forward- and backward-prediction by constraining the agreement between the two directions. This model improves state-of-the-art performance by 9.6% for template-free approaches where the reaction type is unknown.
This paper has not been read by Pith yet.
Forward citations
Cited by 1 Pith paper
-
RETROSPECT: RETROsynthesis via Sequential Prediction, and Chemically Transformed-ranking
RETROSPECT reports 55.00% top-1 and 86.18% top-10 accuracy on USPTO-50K with a ChemAlign Transformer plus LambdaMART reranker reaching 59.4% top-1 on candidate pools using proposal scores and template statistics.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.