Small molecule drug virtual screening method and device based on deep parameter transfer learning

Abstract

The invention provides a small molecule drug virtual screening method and device based on deep parameter transfer learning. The method comprises the steps that molecular fingerprints represent ligandsample characteristics in a data set; feature selection is performed based on group sparse learning to obtain a key substructure; and the activity of the ligand small molecules is predicted based on deep parameter transfer learning. According to the invention, a good deep learning model is trained through similar drug targets rich in training samples, and the deep learning model of a target drug target is initialized by utilizing just learned model parameters according to the hypothesis that the similar drug targets are easy to fit to the similar deep learning model; and finally, the model isoptimized and updated by using limited training samples of the target drug target. The method based on deep parameter transfer learning can be used for trying to solve the problem of insufficient ligand samples in a drug virtual screening training data set, and has potential application value for virtual screening of new targets, understanding of interaction between ligands and the targets and optimization of ligand molecules.

Description

technical field [0001] The present application relates to a method and device for virtual screening of drugs, in particular to a method and device for virtual screening of small molecule drugs based on deep parameter transfer learning. Background technique [0002] In the process of drug development, lead compounds are compounds with certain biological activity and chemical structure obtained through various ways and means, which are used for further structural modification and modification, and are the starting point of modern new drug research. In the process of new drug research, obtaining biologically active lead compounds through compound activity screening is the basis of innovative drug research. [0003] Traditional drug screening requires a lot of manpower and material resources, and there are a series of shortcomings such as long experimental cycle. With the rapid development of computers in the 21st century, virtual drug screening technology has been widely used ...

AI Technical Summary

Problems solved by technology

Therefore, when a small sample is provided, it will be difficult to construct a good virtual screening model due to the lack of sufficient sample size

Images