NEDD: a network embedding based method for predicting drug-disease associations.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug discovery is costly, time-consuming, and high-risk.
• Drug repositioning offers a strategy to reduce costs and time by identifying new uses for existing drugs.
• Accurate identification of novel drug-disease associations requires leveraging diverse information.

Purpose of the Study:

• To propose a novel meta-path-based computational method, NEDD, for predicting drug-disease associations.
• To utilize heterogeneous information for enhanced prediction accuracy in drug repositioning.
• To address the challenge of distinguishing true associations from numerous potential links.

Main Methods:

• Constructed a heterogeneous network integrating drug-drug similarity, disease-disease similarity, and known drug-disease associations.
• Employed meta paths of varying lengths to capture high-order proximity between drugs and diseases.
• Utilized a random forest classifier for predicting novel drug-disease associations based on learned low-dimensional representations.

Main Results:

• The NEDD method demonstrated superior performance in predicting drug-disease associations.
• Experiments were conducted on a gold standard dataset comprising 1933 validated associations.
• NEDD outperformed existing state-of-the-art approaches in prediction accuracy.

Conclusions:

• NEDD is an effective computational approach for drug repositioning.
• The method accurately predicts novel drug-disease associations by exploiting heterogeneous information.
• NEDD offers a promising tool to accelerate drug discovery and development.