A Deep Learning Method, Device, Medium, and Equipment for Predicting Drug Interactions

By combining atomic and molecular-level networks, the relationship information between drug molecules is captured, and the problem of ignoring the internal graph structure information of drug molecules in the prior art is solved, the accuracy of drug interaction prediction is improved, and safe drug prescription design is promoted.

CN114530258BActive Publication Date: 2025-05-27SOUTH CHINA UNIV OF TECH
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202210105604.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-28
Publication Date
2025-05-27
Estimated Expiration
2042-01-28

AI Technical Summary

Technical Problem

Existing deep learning-based drug interaction prediction methods ignore graph structure information inside drug molecules, resulting in insufficient accuracy of prediction results.

Method used

A deep learning drug interaction prediction method is used to capture the relationship information within and between drug molecules through a combination of atomic and molecular networks. The atomic network uses the Transformer model to process the graph structure inside drug molecules, and the molecular network uses a multi-headed attention mechanism to extract the relationship between different drug molecules.

Benefits of technology

It improves the accuracy of drug interaction prediction, helps researchers pre-select drug molecules that may produce interactions, design safer drug prescriptions, and reduces the probability of adverse reactions when combined medication.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114530258B_ABST
    Figure CN114530258B_ABST
Patent Text Reader

Abstract

The present invention provides a deep learning drug interaction prediction method, device, medium and equipment; the method is as follows: obtaining drug molecular information of two drugs to be predicted; an atomic-level network encodes each drug molecular information, captures the interaction information between atoms and chemical bonds, and outputs an encoded drug molecular graph representation z_atom j ; a molecular-level network uses a multi-head attention mechanism to extract the relationships between different drug molecules from each drug molecular graph representation z_atom j respectively, and outputs a molecular graph representation z_mol j ; converting the output molecular graph representations z_mol1 and z_mol2 of the two drugs into a vector, and further obtaining a drug interaction prediction result. This method can solve the problem that edge information cannot be fully considered in the traditional framework, can capture the relationship information between different drug molecules, and thus improve the accuracy of the prediction result.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Small molecule representation learning method based on Transform and enhanced interactive MPNN neural network

    CN113299354A