Scoring multiple features to predict drug disease associations using information fusion and aggregation

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning aims to find new uses for existing drugs, a challenging area in pharmaceutical science.
• Current computational methods often overlook crucial drug/disease features, their importance, and data uncertainty.
• Leveraging diverse data sources and feature fusion can enhance the accuracy of predicting drug-disease interactions.

Purpose of the Study:

• To investigate the impact of drug and disease features on predicting drug-disease interactions.
• To evaluate the effectiveness of data fusion techniques in enhancing prediction accuracy.
• To introduce and validate a novel computational method for drug-disease association prediction.

Main Methods:

• Proposed a novel computational method named scored mean kernel fusion (SMKF).
• Employed a feature fusion approach to create high-level features by combining multiple data sources.
• Systematically integrated drug-drug and drug-disease level features using a scored mean aggregation operator.

Main Results:

• The SMKF method demonstrated superior performance compared to existing approaches.
• Achieved a high area under the curve (AUC) of 0.91.
• Obtained an F-measure of 84.9% and a Matthews correlation coefficient of 70.31% on a gold-standard dataset.

Conclusions:

• The proposed SMKF method effectively predicts drug-disease interactions by integrating diverse features and employing data fusion.
• This approach enhances drug repositioning efforts by identifying novel drug indications with high accuracy.
• SMKF offers a promising computational strategy for pharmaceutical research and drug discovery.