Molecular docking method and system based on migration learning

Abstract

The invention provides a molecular docking method and system based on migration learning. The method comprises: acquiring 3-dimensional space coordinates, van der Waals radius and atom types of a plurality of docking samples; dividing the channel number according to the atomic type of the docking samples, and calculating corresponding values on each grid point according to the 3-dimensional spacecoordinates and the van der Waals radius to obtain multi-channel 3-dimensional grid data; inputting the multi-channel 3-dimensional grid data into the network architecture, outputting the scores of each docking samples and classifying the docking samples according to the scores, so as to train the docking model, wherein for the first trained docking model used for different families of migration learning, different docking models are trained for different families of different docking samples. The invention can automatically train a docking model according to the docking samples, so that the docking model can accurately complete the molecular docking and conform to the practical application scenarios of different families of samples, and improve the accuracy of molecular docking.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a transfer learning-based molecular docking method and a transfer learning-based molecular docking system. Background technique [0002] At present, there are two main difficulties in the field of molecular docking technology, one is computational efficiency, and the other is scoring function. [0003] With the research of GPU general-purpose computing, the problem of computing efficiency has been alleviated to a certain extent, but the scoring problem of judging the quality of docking still needs to be solved. Traditional scoring strategies mainly include evaluation functions based on knowledge, evaluation functions based on experience, and evaluation functions based on molecular force fields, but these scoring strategies have their own defects, and it is generally difficult to effectively learn automatically based on docking samples, so as to give Make a correc...

AI Technical Summary

Problems solved by technology

[0003] With the research of GPU general computing, the problem of computing efficiency has been alleviated to a certain extent, but the problem of scoring the quality of docking still needs to be solved

Traditional scoring strategies mainly include evaluation functions based on knowledge, evaluation functions based on experience, and evaluation functions based on molecular force fields, but these scoring strategies have their own defects, and it is generally difficult to effectively learn automatically based on docking samples, so as to give Make a correct docking situation

In addition, for the actual scene where different samples are docked, there are still cases where the accuracy of docking is low

Images