DRMAHGC

Drug Repositioning by Multi-Aspect Heterogeneous Graph Contrastive Learning and Positive-Fusion Negative Sampling Strategy

Requirements:

• python 3.9.13
• cudatoolkit 11.3.1
• pytorch 1.10.0
• dgl 0.9.0
• networkx 2.8.4
• numpy 1.23.1
• scikit-learn 0.24.2

Data:

The data files needed to run the model, which contain B-dataset, C-dataset and F-dataset.

• DrugFingerprint, DrugGIP: The similarity measurements of drugs to construct the similarity network
• DiseasePS, DiseaseGIP: The similarity measurements of diseases to construct the similarity network
• Protein_sequence, ProteinGIP_Drug, ProteinGIP_Disease: The similarity measurements of proteins to construct the similarity network
• DrugDiseaseAssociationNumber: The known drug disease associations
• DrugProteinAssociationNumber: The known drug protein associations
• ProteinDiseaseAssociationNumber: The known disease protein associations

SGMAE:

• Implementation of SGMAE
• Embedding/: The high-order feature embeddings of drugs and diseases obtained by SGMAE
• Usage: Execute python train_GAE.py

Code:

• data_preprocess.py: Methods of data processing
• metric.py: Metrics calculation
• pos_contrast.py: Get positive and negative samples for contrastive learning
• Contrast.py: Code of graph contrastive learning
• model.py: Model of DRMAHGC
• train_DDA.py: Train the model

Usage:

Execute python train_DDA.py