Heterogeneous graph inference with matrix completion for computational drug repositioning.

Area of Science:

• Computational drug discovery and development
• Bioinformatics and systems biology
• Pharmacology and therapeutics

Background:

• Traditional drug discovery is resource-intensive, necessitating efficient computational approaches.
• Computational drug repositioning accelerates the identification of new uses for existing drugs.
• Existing heterogeneous graph inference methods struggle with data sparsity and suboptimal similarity measures.

Purpose of the Study:

• To develop an accurate and efficient computational drug repositioning method.
• To enhance heterogeneous graph inference by addressing network sparsity and improving similarity metrics.
• To predict potential drug-associated indications for both approved and novel drugs.

Main Methods:

• Implemented a heterogeneous graph inference with matrix completion (HGIMC) method.
• Utilized a bounded matrix completion (BMC) model to prefill missing drug-disease associations.
• Employed Gaussian radial basis function (GRB) for improved drug and disease similarity calculations.

Main Results:

• HGIMC demonstrated superior prediction performance compared to five state-of-the-art methods.
• The method achieved excellent computational efficiency in 10-fold cross-validation and de novo tests.
• Case studies confirmed the practical effectiveness of HGIMC in identifying promising drug indications.

Conclusions:

• HGIMC effectively overcomes limitations of traditional methods in computational drug repositioning.
• The proposed approach enhances accuracy and efficiency in predicting drug-disease associations.
• HGIMC offers a valuable tool for accelerating drug discovery and repurposing efforts.