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

Abstract

In computer-assisted synthesis planning (CASP) programs, providing as many chemical synthetic routes as possible is essential for considering optimal and alternative routes in a chemical reaction network. As the majority of CASP programs have been designed to provide one or a few optimal routes, it is likely that the desired one will not be included. To avoid this, an exact algorithm that lists possible synthetic routes within the chemical reaction network is required, alongside a recommendation of synthetic routes that meet specified criteria based on the chemist's objectives. Herein, we propose a chemical-reaction-network-based synthetic route recommendation framework called "Com-pRet" with a mathematically guaranteed enumeration algorithm. In a preliminary experiment, CompRet was shown to successfully provide alternative routes for a known antihistaminic drug, cetirizine. CompRet is expected to promote desirable enumeration-based chemical synthesis searches and aid the development of an interactive CASP framework for chemists.

PaRoutes: a framework for benchmarking retrosynthesis route predictions

“…Although speed and number of solved targets are interesting metrics, they tell nothing about the quality of the predictions. Sometimes the quality of the routes has been quantified as the length of the routes [21,26], the mean chemical complexity [27], or a metric based on the priors from the one-step model [21]. Quantifying the quality of a route by simply the route length is particularly misleading as it is very easy to envisage a shorter alternative to a route by for instance removing all protection chemistry thus rendering the route chemically infeasible.…”

“…Quantifying the quality of a route by simply the route length is particularly misleading as it is very easy to envisage a shorter alternative to a route by for instance removing all protection chemistry thus rendering the route chemically infeasible. An interesting approach taken by Shibukawa et al [27] was to sort the molecules in the routes by molecule weight and then compute the pairwise Tanimoto distances of a molecular fingerprint. Naturally, the sorting of the molecules destroys the order of the reactions in the route.…”

“…The proof-number and disproof-number of each node, which is the number of children nodes necessary to prove and disprove a node, respectively, are used to guide the search. PNS-based retrosynthesis algorithms utilize these numbers differently and also add additional heuristics to increase the efficiency of the search [20,26,27]. We implemented a depth-first PNS (DFPN) algorithm in the AiZynthFinder software that combines ideas from two different algorithms [26,27].…”

“…PNS-based retrosynthesis algorithms utilize these numbers differently and also add additional heuristics to increase the efficiency of the search [20,26,27]. We implemented a depth-first PNS (DFPN) algorithm in the AiZynthFinder software that combines ideas from two different algorithms [26,27]. In MCTS, the retrosynthesis is carried out in a different type of direct graph, where each node is a super-node consisting of one or more molecules that potentially can be extended.…”

Fast prediction of distances between synthetic routes with deep learning

“…A search algorithm should not only return the synthetic route that is most likely to succeed, but additionally a set of diverse routes that a chemist can analyze further if the top-ranked route is not adequate. The pruning of the identified routes to create a diverse set of routes can be done as part of the search algorithm itself [11] or as a post-processing step [12].…”