Using Machine Learning To Predict Suitable Conditions for Organic Reactions

Abstract

Reaction condition recommendation is an essential element for the realization of computer-assisted synthetic planning. Accurate suggestions of reaction conditions are required for experimental validation and can have a significant effect on the success or failure of an attempted transformation. However, de novo condition recommendation remains a challenging and under-explored problem and relies heavily on chemists’ knowledge and experience. In this work, we develop a neural-network model to predict the chemical context (catalyst(s), solvent(s), reagent(s)), as well as the temperature most suitable for any particular organic reaction. Trained on ∼10 million examples from Reaxys, the model is able to propose conditions where a close match to the recorded catalyst, solvent, and reagent is found within the top-10 predictions 69.6% of the time, with top-10 accuracies for individual species reaching 80–90%. Temperature is accurately predicted within ±20 °C from the recorded temperature in 60–70% of test cases, with higher accuracy for cases with correct chemical context predictions. The utility of the model is illustrated through several examples spanning a range of common reaction classes. We also demonstrate that the model implicitly learns a continuous numerical embedding of solvent and reagent species that captures their functional similarity.

Datasets and their influence on the development of computer assisted synthesis planning tools in the pharmaceutical domain

“…This task is le to separate models that have not been implemented in this study, that attempt to predict conditions for a queried set of substrates and a given transformation. 36 The under-prediction of retrosynthetic routes to compounds that were experimentally obtained in the 'bespoke' libraries, raises questions as to the coverage of the reaction space covered by the templates, and the ability of the policy network to prioritize suitable templates. Fig.…”

Machine Learning Applied to Zeolite Synthesis: The Missing Link for Realizing High-Throughput Discovery

“…We need to develop improved descriptors that can capture the properties we intend to model in more effective ways. 53 including the unsupervised selection of reaction conditions, 54 catalysts, and reagents, to deploy AI-driven experimental platforms including full automation for highly parallelized OSDA synthesis. OSDA design could benefit from cutting-edge algorithms, such as generative adversarial networks and reinforcement learning already used in the design of biological compounds, in which new molecules with specific physicochemical features are produced using a punishment/reward system analogous to that in psychological conditioning.…”

Challenges and Directions for Green Chemical Engineering—Role of Nanoscale Materials

“…Neural networks are popular ML models, which have been proved as general approximators of any non-linear function with appropriate setting of network architecture, and have been used successfully for prediction reaction results. 122,123,124 Data-driven chemical synthesis planning is another new trend, to seek help of machine learning. 125,126 Reaction rules are extracted from large reaction corpora (such as the United States Patent and Trademark Office (USPTO), Reaxys, and SciFinder databases).…”