AiZynthFinder: A Fast Robust and Flexible Open-Source Software for Retrosynthetic Planning

Abstract

We present the open-source AiZynthFinder software that can be readily used in retrosynthetic planning. The algorithm is based on a Monte Carlo tree search that recursively breaks down a molecule to purchasable precursors. The tree search is guided by an artificial neural network policy that suggests possible precursors by utilizing a library of known reaction templates. The software is fast and can typically find a solution in less than 10 seconds and perform a complete search in less than 1 minute. Moreover, the writing of the code was guided by a range of software engineering principles such as automatic testing, system design and continuous integration leading to robust software. The object-oriented design makes the software very flexible and can straightforwardly be extended to support a range of new features. Finally, the software is clearly documented and should be easy to get started with. The software is available at http://www.github.com/MolecularAI/aizynthfinder.

Retrosynthetic accessibility score (RAscore) – rapid machine learned synthesizability classification from AI driven retrosynthetic planning

“…AiZynthFindera tool for computer aided synthesis planning AiZynthFinder is a template-based retrosynthetic planning tool based on the methodology of Segler and Waller. 4,16 It consists of a neural network policy, which determines which reaction to use at a given retrosynthetic step, with Monte-Carlo tree search, as reported in our previous studies. 24 The code, data, and models are open source and available to the public: https:// github.com/MolecularAI/AiZynthFinder.…”

“…[11][12][13] CASP has emerged as a method by which compounds can be ltered in the VS workow, and during optimization cycles throughout the generative modelling process. Several recent CASP tools have been developed which may be used for these purposes, including but not limited to: Synthia (formerly Chematica), 14 ICSYNTH, 15 ASKCOS, 5 AiZynthFinder, 16 and IBM RXN. 17 These can be used at two potential stages of the generation process, either to bias the generation process or as a post hoc lter aer the molecules have been generated.…”

“…To tackle the challenge of screening large compound libraries with synthesizability considerations, existing scores include the synthetic accessibility score (SAscore), synthetic complexity score (SCscore), and synthetic Bayesian accessibility (SYBA). 16,[21][22][23] The SAscore and SYBA are estimations of synthetic feasibility based on the occurrence of molecular fragments in public databases, whereas SCscore is learned from a reaction corpus, with the underlying assumption that products are more complex than their constituent reactants.…”

“…Herein, we propose the retrosynthetic accessibility score (RAscore) that enables rapid estimation of synthetic feasibility as determined from the predictions of CASP, in this case AiZynth-Finder. 16 We investigate a machine learning classier for retrosynthetic accessibility (RA) assessment called RAscore, trained on the outcomes generated from AiZynthFinder, which we have shown can increase the speed at which synthetic accessibility can be estimated, and separate compounds for which retrosynthetic routes can be found by AiZynthFinder. This is an improvement that adds value to existing synthesis scores, and when used in combination with the previous scores, the RAscore should enable prescreening of compounds that can be later subjected to full retrosynthetic analysis.…”

From Big Data to Artificial Intelligence: chemoinformatics meets new challenges

“…The benchmarking studies [20,22] show that such methods can achieve similar or better performances compared to traditional methods for classical tasks such as QSAR [24] but at the same time allow for intuitive interpretation of models [21]. Moreover, they can be used to address very different tasks, such as the aforementioned generation of molecules with desired properties or/and the prediction of single step (retro) synthesis [25,26], or even complete retro-synthesis [27,28] that could not be achieved with traditional methods. All these approaches are part of the emerging area of AI, which is going to drive the future of chemoinformatics.…”