Intelligent molecular design method based on auto-encoder and third-order graph convolution

Abstract

The invention provides an intelligent molecular design method based on an auto-encoder and third-order graph convolution. The method comprises the following steps: 1, enabling an encoder to express amedicine molecule in a molecular graph form, and enabling each molecular graph to be decomposed into a corresponding third-order substructure; 2, performing graph convolution on the third-order substructure by the encoder, and outputting an implicit vector of a molecule to acquire an implicit vector space; 3, training a decoder; and 4, enabling the decoder to find an implicit vector similar to themedicine molecule in the implicit vector space, and decoding the implicit vector to obtain a new medicine molecule. The method disclosed by the invention has the advantages that the three-order substructure of a molecule is subjected to spatial graph convolution by combining an auto-encoder and a graph convolution network method to find out an atom corresponding substructure, namely a functionalgroup, with high contribution degree to the molecule.

Description

technical field [0001] The invention belongs to the field of molecular intelligent design methods, in particular to a molecular intelligent design method based on an autoencoder and three-order graph convolution. Background technique [0002] At present, Autoencoder is a specific neural network structure whose purpose is to map input information to a lower-dimensional space and generate encoded hidden vectors containing important features. This part is called an encoder. It can be represented by the function h=f(x), and then use the decoder to reconstruct the hidden vector into a result that can restore the original input as much as possible, using the function r=g(h). Our goal is to make g(f(x))=x as much as possible. Of course, it is meaningless to simply copy the input to the output. We need to add certain restrictions to make our model learn from the data more important. Characteristics. [0003] Attention models have been widely used in various fields of deep learning...

AI Technical Summary

Problems solved by technology

[0007] However, the generation of drug molecules mostly relies on the experience of biochemists, which has great subjectivity and limitations. At present, in the work of using deep learning to generate molecules, the functional groups of molecules have not been well excavated and utilized, and the functional groups of molecules For new drugs, play a decisive role

Images