基于深度图神经网络的药物-疾病关联预测方法
By constructing a drug-disease similarity matrix using a deep graph neural network and employing random walks and graph neural networks for supervised learning, the problem of insufficient prediction accuracy in drug relocation is solved, achieving more efficient drug-disease association prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEAST FORESTRY UNIV
- Filing Date
- 2023-03-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing drug relocation methods suffer from insufficient prediction accuracy when dealing with highly sparse experimental data, and the representation of drug and disease features is inadequate, making it impossible to effectively mine potential feature information.
A deep graph neural network-based approach is adopted. A similarity matrix between drugs and diseases is constructed using a Gaussian kernel function. Supervised learning is performed using random walks and graph neural networks to extract feature representations of drugs and diseases. Feature extraction is then performed through nonlinear fully connected layers, and finally, a prediction model is constructed.
It improves the accuracy and robustness of drug-disease association prediction, reduces data sparsity, enhances the model's generalization ability, and can better obtain information on the interaction between drugs and diseases.
Smart Images
Figure CN116343909B_ABST