Bayesian Algorithm for Retrosynthesis

Abstract

The identification of synthetic routes that end with the desired product is considered an inherently time-consuming process that is largely dependent on expert knowledge regarding a limited proportion of the entire reaction space. At present, emerging machine learning technologies are reformulating the process of retrosynthetic planning. This study aimed to discover synthetic routes backwardly from a given desired molecule to commercially available compounds. The problem is reduced to a combinatorial optimization task with the solution space subject to the combinatorial complexity of all possible pairs of purchasable reactants. We address this issue within the framework of Bayesian inference and computation. The workflow consists of the training of a deep neural network, which is used to forwardly predict a product of the given reactants with a high level of accuracy, followed by inversion of the forward model into the backward one via Bayes’ law of conditional probability. Using the backward model, a diverse set of highly probable reaction sequences ending with a given synthetic target is exhaustively explored using a Monte Carlo search algorithm. With a forward model prediction accuracy of approximately 87%, the Bayesian retrosynthesis algorithm successfully rediscovered 81.8 and 33.3% of known synthetic routes of one-step and two-step reactions, respectively, with top-10 accuracy. Remarkably, the Monte Carlo algorithm, which was specifically designed for the presence of multiple diverse routes, often revealed a ranked list of hundreds of reaction routes to the same synthetic target. We also investigated the potential applicability of such diverse candidates based on expert knowledge of synthetic organic chemistry.

Scenario prediction of public health emergencies using infectious disease dynamics model and dynamic Bayes

“…For the cause set H of causality in the Bayesian network, the result is expressed as I , then and . In the Bayesian network reasoning process, the reason set for a given node is combined with its result independently [16] . Therefore, the joint probability of all nodes represented by the Bayesian network can be expressed as Eq.…”

Substructure-based neural machine translation for retrosynthetic prediction

“…We compared the prediction accuracy of our approach with other retrosynthetic prediction methods without considering reaction class labels because no prior reaction class information was provided to our model. Several recent reports summarized the prediction accuracy of various models [37,62]. According to reproduced results presented by Lin et al [37], Top-1 accuracy ranges from 28.3% (Liu et al [34] LSTM model over the USPTO 50 K dataset) to 54.1% (Transformer model over the USPTO MIT dataset by Lin et al [37]).…”

“…This is especially true assuming that several possible synthetic routes are available for the forward reaction. It is worth noting that the content of the dataset used in the reverse mapping, could also be responsible for the network's behavior [62].…”

Substructure-based Neural Machine Translation for Retrosynthetic Prediction

“…Several recent reports summarized the prediction accuracy of various models. 37,61 According to reproduced results presented by Lin et al, 37 top-1 accuracy ranges from 28.3% (Liu et al 34 To identify how our model learns the grammar of chemical reactions, the evolution of The quality of bad predictions (ca. 5% of the validation set) did not improve probably due to the insufficient information, complexity, and noise contained in the data.…”

“…It is worth noting that the content of the dataset used in the reverse mapping, could also be responsible for the network's behavior. 61 Mapping a reactant from a reactant domain to a product domain and then reversing it does not necessarily produce the original reactant considering the level of abstraction used to describe the molecules in our dataset. There is a chance that the presence of one-to-many mappings from a product to a reactant domain may create confusion during the learning process.…”