Predicting drug-disease associations via sigmoid kernel-based convolutional neural networks

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Computational drug repositioning is crucial for identifying novel drug-disease associations efficiently.
• Deep learning advancements have significantly enhanced data mining capabilities in bioinformatics.
• There is a growing need for advanced computational tools in drug development.

Purpose of the Study:

• To develop an effective computational method for predicting drug-disease associations.
• To leverage deep learning for improved drug repositioning strategies.
• To enhance the discovery of potential new indications for existing drugs.

Main Methods:

• Proposed a novel computational method combining Sigmoid Kernel and Convolutional Neural Network (SKCNN).
• Constructed drug similarity using sigmoid and structural metrics; disease similarity using sigmoid and semantic metrics.
• Employed SKCNN to learn hidden representations and a random forest classifier for label prediction.

Main Results:

• The proposed SKCNN method demonstrated improved prediction accuracy compared to existing state-of-the-art approaches.
• Experimental evaluations confirmed the superior performance of SKCNN.
• Case studies validated the method's effectiveness in discovering potential drug indications.

Conclusions:

• SKCNN is an effective predictive model for identifying new drug-disease associations.
• The method offers a valuable tool for computational drug repositioning.
• SKCNN enhances the discovery of potential therapeutic uses for drugs.