Energy-based View of Retrosynthesis

14 Jul 2020  ·  Ruoxi Sun, Hanjun Dai, Li Li, Steven Kearnes, Bo Dai ·

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.

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Single-step retrosynthesis USPTO-50k Dual-TF (reaction class unknown) Top-1 accuracy 53.6 # 21
Top-3 accuracy 70.7 # 22
Top-5 accuracy 74.6 # 24
Top-10 accuracy 77.0 # 25
Single-step retrosynthesis USPTO-50k Dual-TF (reaction class as prior) Top-1 accuracy 65.7 # 6
Top-3 accuracy 81.9 # 9
Top-5 accuracy 84.7 # 12
Top-10 accuracy 85.9 # 19

Methods