Predicting retrosynthetic pathways using transformer-based models and a hyper-graph exploration strategy

Abstract

We present an extension of our Molecular Transformer model combined with a hyper-graph exploration strategy for automatic retrosynthesis route planning without human intervention. The single-step retrosynthetic model sets a new state of the art for predicting reactants as well as reagents, solvents and catalysts for each retrosynthetic step. We introduce four metrics (coverage, class diversity, round-trip accuracy and Jensen-Shannon divergence) to evaluate the single-step retrosynthetic models, using the forward prediction and a reaction classification model always based on the transformer architecture. The hypergraph is constructed on the fly, and the nodes are filtered and further expanded based on a Bayesian-like probability. We critically assessed the end-to-end framework with several retrosynthesis examples from literature and academic exams. Overall, the frameworks have an excellent performance with few weaknesses related to the training data. The use of the introduced metrics opens up the possibility to optimize entire retrosynthetic frameworks by focusing on the performance of the singlestep model only.

CompRet: a comprehensive recommendation framework for chemical synthesis planning with algorithmic enumeration

“…Although the exponential growth is essentially difficult to overcome, parallel tree search algorithms [51] have the potential to hand this problem. In-scope filter [18] or other graph pruning algorithms [21] will also provide solutions. As shown in the demonstration of chemical reaction network constructions, the explored synthetic routes depend on the template set and maximum depth md.…”

“…While expert systems and knowledge-based programs were the primary focus of CASP during the early stages [4][5][6][7][8], recent breakthroughs in the field of deep learning and widespread availability of reaction datasets have accelerated its development [9][10][11][12][13][14][15][16][17]. In particular, data-driven approaches have received attention across research fields [18][19][20][21]. These approaches for multi-step synthesis planning have shown outstanding performance at every stage, and more recently, they have provided realistic and preferable synthetic routes.…”

Illuminating elite patches of chemical space

“…Some attention should also be drawn to the exciting prospect of steering GB-EPI by direct experimental feedback. Through active learning 55a small-data alternative to deep learningand graph-based retrosynthesis, 56,57 molecules proposed by GB-EPI could be selected for in vitro synthesis and analysis. § The experimental results could then be used to update the tness model.…”

A Recommender System for Inverse Design of Polycarbonates and Polyesters

“…Evaluation of the experimental conditions by means of ML/DL has been gaining momentum. The spectrum of this effort is quite broad, from the experiments on AI predicting organic synthesis, 12 to retrosynthetic approaches for organic molecules [44][45][46] to generative modeling of synthetic conditions for inorganics and mixed materials trained from the literature data. 47,48 Experimentproperty models require adequate information about capabilities of the available experimental platform.…”