Drug sensitivity prediction method and device based on transfer learning and graph neural network

Abstract

The invention discloses a drug sensitivity prediction method and device based on transfer learning and a graph neural network, and belongs to the technical field of drug sensitivity detection and evaluation. The method specifically comprises the steps: performing parameter optimization on a semi-inhibitory concentration prediction module through first sample data containing the semi-inhibitory concentration of a drug to a cell line; migrating the network parameters of a feature extraction unit of a semi-inhibitory concentration prediction module to a feature extraction unit of a drug sensitivity prediction module through migration learning, and on this basis, performing parameter optimization on the drug sensitivity prediction module through second sample data containing drug sensitivity category data; in this way, cross-dataset feature migration is effectively achieved, the defect that the sample size of the datasets is insufficient is overcome, the drug sensitivity prediction module is optimized by means of the datasets with large differences, the drug sensitivity prediction module can effectively extract gene features and drug features of individuals, and the drug sensitivity is predicted with high accuracy.

Description

technical field [0001] The invention belongs to the technical field of drug sensitivity detection and evaluation, and in particular relates to a drug sensitivity prediction method and device based on transfer learning and graph neural network. Background technique [0002] With the development of deep learning and artificial intelligence technology, it has shown great potential in various fields and application scenarios. In biomedicine, people are also actively trying to use deep learning and artificial intelligence technology to automatically learn individual characteristics, thereby promoting the development of precise diagnosis and precise treatment. Among them, how to predict the efficacy of individual drugs based on the biological information of each individual and the structure and properties of a specific drug is an important issue in the field of biomedical research. However, the application of deep learning in bioinformatics often faces problems such as uneven dat...

AI Technical Summary

Problems solved by technology

[0005] In view of the above, the purpose of the present invention is to provide a drug sensitivity prediction method and device based on transfer learning and graph neural network, which uses transfer learning strategy to transfer learning data with asymmetric information, and at the same time cooperates with graph neural network to extract data Features to solve the problem of low robustness of the drug sensitivity prediction model and difficulty in accurately predicting drug sensitivity due to the small sample size and complex feature extraction problems

Images