Asymmetric drug interaction prediction method based on diffusion graph attention network

Abstract

The present invention provides a method for predicting asymmetric drug interactions based on a diffusion graph attention network, which relates to the technical field of drug action prediction. The method uses the chemical structure characteristics of drug molecules to construct a directed graph network, and extracts features from the drug application and reception perspectives respectively through a bidirectional graph attention network, thereby effectively characterizing the asymmetry between drugs. Furthermore, a diffusion model is introduced to perform noise injection and denoising on the graph structure, which significantly enhances the model's adaptability to sparse data and the robustness of feature extraction. Finally, multimodal features are fused through a deep neural network to achieve high-precision prediction of drug interactions. Not only does it surpass existing technologies in prediction accuracy, it also significantly reduces dependence on labeled data, greatly improving the feasibility and generalization ability of the model in practical applications, and providing a new technical path for the safety assessment of drug combination therapy and precision medicine research.

Images