HNet-DNN: Inferring New Drug-Disease Associations with Deep Neural Network Based on Heterogeneous Network Features

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug discovery is costly and time-consuming.
• Drug repositioning offers an economical alternative.
• High-throughput data necessitates advanced computational methods for identifying new drug indications.

Purpose of the Study:

• To introduce HNet-DNN, a deep neural network (DNN) model for predicting novel drug-disease associations.
• To leverage drug-disease heterogeneous networks for improved prediction accuracy.
• To provide an efficient computational approach for drug repositioning.

Main Methods:

• Constructed drug-drug and disease-disease similarity networks from raw features.
• Built a drug-disease heterogeneous network integrating known associations.
• Extracted topological features from the heterogeneous network to train a DNN model.
• Utilized deep neural networks (DNN) for predictive modeling.

Main Results:

• HNet-DNN demonstrated superior performance in predicting drug-disease associations.
• The method effectively utilized heterogeneous network features.
• Achieved state-of-the-art results compared to existing approaches.
• Successfully alleviated the overfitting problem in predictions.

Conclusions:

• HNet-DNN is a powerful tool for drug repositioning.
• The approach enhances the efficiency of identifying novel drug indications.
• Case studies validated the method's effectiveness in predicting new drug-disease links.