Ionic liquid type antibiotic drug property prediction method based on transfer learning and graph neural network and high-throughput screening platform

Abstract

The invention relates to an ionic liquid antibiotic drug property prediction method based on transfer learning and a graph neural network and a high-throughput screening platform. The method comprises the following steps: firstly, collecting minimum inhibitory concentrations of organic molecules and ionic liquid to germs from a public database and literatures, and respectively taking the minimum inhibitory concentrations as a data set for pre-training and a data set for final training; pre-training the single image input model by adopting data of organic molecules; pre-training model parameters are transplanted into the multi-graph input model through a transfer learning method; training the multi-graph input model by using the data of the ionic liquid to obtain an ionic liquid antibacterial performance prediction module; a theoretical structure library of the ionic liquid is constructed; and performing antibacterial property prediction on the ionic liquid in the theoretical structure library through a prediction module, and finally screening out an ionic liquid structure with excellent antibacterial property. According to the method, the problems of small data volume and deep learning prediction of a complex ion structure-drug relationship are solved, and high-throughput screening of novel ionic liquid antibiotics is realized.

Description

technical field [0001] The invention relates to a drug property prediction method and a high-throughput screening platform for ionic liquid antibiotics based on migration learning and graph neural network, which belongs to the field of biomedical application of artificial intelligence, and is especially aimed at a computational new drug discovery method. Background technique [0002] Due to their unique structure and physicochemical properties, ionic liquids (ILs) have been widely used in the research and application of solvents, catalysis, and medicine. However, the huge number of ionic liquids cannot synthesize all possible cations and anions. Therefore, it is of great significance to predict the physicochemical and pharmaceutical properties of ILs based on the structure. Currently, the prediction of the properties of ILs by computational chemistry (molecular dynamics, quantification, etc.) methods has achieved rapid development. However, these methods are computationall...

AI Technical Summary

Problems solved by technology

[0004] The problem to be solved by the present invention is to provide a deep learning prediction method and a high-throughput screening platform for end-to-end learning and good scalability in view of the difficulties of the small amount of antibacterial data and the complex structure-drug property relationship of ionic liquids

Images