Graph feature extraction method, lipid-water partition coefficient prediction method and graph feature extraction model

Abstract

The invention discloses a graph feature extraction method, a lipid-water partition coefficient prediction method and a graph feature extraction model. The graph feature extraction method comprises thesteps: obtaining a to-be-extracted feature graph which consists of a plurality of nodes and edges connected with the nodes with an association relation; inputting the to-be-extracted feature graph into a graph feature extraction model for feature extraction to obtain features of each node, the graph feature extraction model comprising a plurality of convolution layers and GRU network layers, theconvolution layers and the GRU network layers being arranged at intervals, and performing feature fusion on nodes with the association relation through the GRU network layers; and inputting the features of each node output by the last convolution layer into the merging layer for feature fusion to obtain the features of the to-be-extracted feature graph. According to the method, the GRU network layers are used for fusing the feature information of the nodes with the association relation during each convolution operation, so that the expression capability of the network is better, the interaction between the nodes in the graph can be expressed better, and the early-stage extraction workload is reduced.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a graph feature extraction method, a fat-water partition coefficient prediction method and a graph feature extraction model. Background technique [0002] Fat-water partition coefficient log p =Concentration of the substance in octanol / concentration of the substance in water, this index is an important reference element in drug design, and it affects the absorption and operation of the drug in the body. Although this index can be measured by simple experiments, it is unrealistic to test a large number of candidate small molecules in the early virtual screening stage of drug design. At this time, pharmaceutical and chemical experts often use software calculation methods to calculate log p Perform a coarse sieve. [0003] In related technologies, machine learning models are usually used to analyze the log of small molecules p Prediction, but feature extraction re...

AI Technical Summary

Problems solved by technology

[0004] Therefore, the technical problem to be solved in the present invention is to overcome the defect that feature extraction in the prior art requires a lot of preliminary work and a large workload, thereby providing a graph feature extraction, fat-water partition coefficient prediction method and graph feature extraction model

Images