Method for predicting protein and ligand molecule binding free energy based on convolutional neural network

Abstract

The invention provides a method for predicting protein and ligand molecule binding free energy based on a convolutional neural network. The method comprises the following steps: (1) obtaining a high-quality protein-ligand compound crystal structure from a PDB database, collecting a small molecular structure from literature, and performing 3D conformation preparation; (2) performing encoding and descriptor calculation on the protein and the small molecules, and processing the protein into a 3D image; (3) using the 3D image of the protein as an input, and designing a convolutional neural networkarchitecture which comprises an input layer, a hidden layer and an output layer; (4) carrying out feature extraction and fine tuning on the pre-trained model so as to be applied to a new data set; and (5) after all the tasks are completed, calling Panda and Matplotlib data analysis libraries, and directly drawing an analysis result curve in Jupyter. According to the method, rapid and accurate protein-ligand binding free energy calculation is achieved, compared with an experimental value, errors are small, and data analysis and visualization can be automatically conducted on the result.

Description

technical field [0001] The invention belongs to the field of drug research and development based on artificial intelligence, specifically a method for predicting the binding free energy of proteins and ligand molecules based on convolutional neural networks, and realizes the precise prediction of the binding freedom of receptors and drug molecules by applying convolutional neural networks (CNN) It can be applied to the design and development of new drugs. Background technique [0002] New drug design and development is a creative and exploratory research work. Drug molecular design is based on rational strategies and scientific planning to construct new molecular entities with expected pharmacological activities. Molecular design is to gradually optimize active compounds into compounds that are safe, effective, controllable and easy to obtain in the human body, and meet the requirements for multidimensional properties of drugs during the process of changing and modifying st...

AI Technical Summary

Problems solved by technology

For example, the FEP method faces problems such as the preparation of the simulation system, the processing of the solvent model, the selection of the force field, and the calculation cost, which makes it difficult to apply on a large scale in practice.

The scoring function method is mainly obtained by training and fitting on some data sets with very different chemical structures, and it is difficult to generalize to homologue ligands with small structural differences, so it has limited application in the lead optimization scenario.

Images