Asymmetric catalytic reaction enantioselectivity prediction method based on deep learning

Abstract

The invention discloses an asymmetric catalytic reaction enantioselectivity prediction method based on deep learning. The method comprises the following steps: firstly, acquiring and sorting asymmetric catalytic reaction data in which isocyano acetate participates, and designing a model training set and a sample external test set; calculating and processing molecular descriptors of reaction related compounds, summarizing the molecular descriptors and reaction conditions into a group of feature vectors, and inputting the feature vectors into a model; respectively constructing a deep neural network regression model and a convolutional neural network regression model based on the training set, and optimizing hyper-parameters of the deep neural network regression model and the convolutional neural network regression model, so as to obtain a model capable of accurately predicting the reaction enantioselectivity of the training set; and utilizing the optimal neural network model to predict the enantioselectivity of the in-vitro reaction of the sample, and checking the mobility of the model. The result shows that the model can accurately predict the enantioselectivity of the in-vitro reaction of the sample, and the robustness and mobility of the model are further verified.

Description

technical field [0001] The invention relates to the interdisciplinary field of computer science and chemical organic synthesis, in particular to a method for predicting enantioselectivity of asymmetric catalytic reactions based on deep learning. Background technique [0002] The enantioselectivity of asymmetric catalytic reactions has great reference value for the efficient and precise synthesis of target chiral molecules, and mastering the law of enantioselectivity in reactions has a huge role in promoting the development of new drugs. However, the development of traditional asymmetric catalytic reaction systems is extremely dependent on personal experience, and the target reaction is achieved through a wide range of screening and condition optimization, which is time-consuming, laborious and has a low success rate. Artificial intelligence technology can make predictions and judgments by learning the deep information hidden in the data and mining the internal correlation. ...

AI Technical Summary

Problems solved by technology

However, these works still have some limitations: (1) some of the work involves a single type of reaction, the number of samples is small, the model results are only applicable to this type of reaction, and the transferability of the model is poor; (2) the overall method of some work It is more complicated, and it is more difficult to reproduce and migrate the model

Images