Method for predicting dissociation constant of small molecule compounds based on graph convolutional neural network

Abstract

The invention relates to a method for predicting a dissociation constant of a small molecule compound based on a graph convolutional neural network, which is different from a traditional prediction method based on molecular fingerprints, and automatically learns a chemical mode related to the dissociation constant through the graph convolutional neural network to construct a prediction model. The dissociation constants of 16,000,000 compounds in the ChEMBL database are used for training the model. In addition, in order to quickly determine the dissociation center in each compound, a substructure database containing 144 SMARTS templates is constructed on the basis of a ChEMBL database. When a new compound is predicted, firstly, the SMARTS template library is used for matching dissociation centers in the compound, and then the dissociation constants of the dissociation centers are predicted in sequence. By utilizing the method, the dissociation constants of the compounds can be quickly and accurately predicted, and the efficiency of drug design and virtual screening is improved.

Description

technical field [0001] The invention relates to the field of computer-aided drug design, which is a method for predicting dissociation constants of small molecule compounds by computer based on graph convolutional neural network, which can be applied in the field of drug virtual screening and drug molecular design, and shortens the cycle of drug research and development. Background technique [0002] Dissociation constant is an important physicochemical property of small molecule drugs. It has been reported that about 79% of orally administered drugs contain at least one ionizable group. It determines the activity, solubility, lipophilicity, bioaccumulation and toxicity of drug molecules. At the same time, it will also affect the pharmacokinetics and biochemical properties of drug molecules. [0003] At present, there are three main methods for predicting dissociation constants. One is the linear free energy method, which establishes a linear equation based on the substitu...

AI Technical Summary

Problems solved by technology

At the same time, it will also affect the pharmacokinetics and biochemical properties of drug molecules

[0003] At present, there are three main methods for predicting dissociation constants. One is the linear free energy method, which establishes a linear equation based on the substituent effect, but this type of method needs to extract parameters based on experience, and establish a parameter library and a matrix library; The QSPR method uses machine learning to model the descriptors of the chemical environment around the dissociation center. The descriptors determine the quality of this type of method; there is also a computing method based on quantum chemistry, which requires a lot of computing power. , the calculation speed is slow

Images