Compound property prediction method, device, computer equipment and readable storage medium

Abstract

The embodiment of the present application discloses a compound property prediction method, device, computer equipment and readable storage medium, involving computer vision technology of artificial intelligence; specifically, the chemical structure information of the target compound can be obtained, the chemical structure information includes atoms and chemical bonds ; Generate a chemical structure graph corresponding to the chemical structure information according to the chemical structure information, the chemical structure graph includes the nodes corresponding to the atoms and the edges corresponding to the chemical bonds; construct the original node features of the nodes and the original edge features of the edges; according to the original node features of the nodes, The original edge characteristics of the edge, multiple rounds of message propagation on the edge, and the propagation state information of the edge after multiple rounds of message propagation are obtained; the target feature of the edge is obtained according to the propagation state information; the properties of the target compound are predicted according to the target feature of the edge, And output the property prediction results of the target compound. This program greatly improves the accuracy of compound property prediction.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a compound property prediction method, device, computer equipment and readable storage medium. Background technique [0002] In drug analysis, how to judge or predict some properties of substances such as chemical and biological properties based on the structure of compounds such as proteins, such as toxicity, solubility, carcinogenicity, etc., is a very important task. [0003] The current way to predict the properties of compounds is to use artificial intelligence (AI, Artificial Intelligence) to achieve. For example, tasks such as property prediction can be modeled as classification problems in AI machine learning. Specifically, the structural formula of a compound such as a protein can be converted into a vector in the Euclidean space by means of representation learning, and then it can be classified by a neural network such as a convolutional neural ...

AI Technical Summary

Problems solved by technology

However, compounds such as protein chemical structures do not have such a lattice structure, resulting in lower accuracy of property prediction

Images