BGMSDDA: a bipartite graph diffusion algorithm with multiple similarity integration for drug-disease association...

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning identifies new uses for existing drugs, saving time and cost.
• Current methods struggle with sparse drug-disease association data.
• A need exists for methods to mine similar drug and disease features.

Purpose of the Study:

• To develop a novel algorithm for predicting drug-disease associations.
• To address data sparsity in drug repositioning.
• To integrate multiple similarity measures for improved prediction accuracy.

Main Methods:

• Developed the bipartite graph diffusion algorithm with multiple similarity integration (BGMSDDA).
• Reconstructed the drug-disease association matrix using the weight K nearest known neighbors (WKNKN) algorithm.
• Extracted drug and disease features by integrating linear neighborhood and Gaussian kernel similarities.
• Employed bipartite graph diffusion for inferring novel drug-disease associations.

Main Results:

• BGMSDDA demonstrated excellent performance in 10-fold cross-validation experiments.
• Achieved high AUC values (0.939 on Fdataset, 0.954 on Cdataset) and AUPR values (0.466 on Fdataset, 0.565 on Cdataset).
• Case studies validated the predictive accuracy of associated diseases for selected drugs.

Conclusions:

• BGMSDDA effectively predicts drug-disease associations, overcoming data sparsity issues.
• The algorithm shows significant potential for accelerating drug repositioning and development.
• This method offers a valuable tool for discovering new therapeutic indications for existing drugs.