Abstract:A fundamental problem in computational chemistry is to find a set ofreactants to synthesize a target molecule, a.k.a. retrosynthesis prediction.Existing state-of-the-art methods rely on matching the target molecule with alarge set of reaction templates, which are very computationally expensive andalso suffer from the problem of coverage. In this paper, we propose a noveltemplate-free approach called G2Gs by transforming a target molecular graphinto a set of reactant molecular graphs. G2Gs first splits the target moleculargraph into a set of synthons by identifying the reaction centers, and thentranslates the synthons to the final reactant graphs via a variational graphtranslation framework. Experimental results show that G2Gs significantlyoutperforms existing template-free approaches by up to 63top-1 accuracy and achieves a performance close to that of state-of-the-arttemplate based approaches, but does not require domain knowledge and is muchmore scalable.

A Graph to Graphs Framework for Retrosynthesis Prediction