Human Microbe-Disease Association Prediction With Graph Regularized Non-Negative Matrix Factorization

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Microbes are intricately linked to human health, influencing disease prevention, diagnosis, and treatment.
• Computational models offer efficient and cost-effective alternatives to traditional experiments for predicting microbe-disease associations.

Purpose of the Study:

• To develop a novel computational model, Graph Regularized Non-negative Matrix Factorization for Human Microbe-Disease Association prediction (GRNMFHMDA), for identifying potential microbe-disease associations.
• To evaluate the predictive performance and reliability of the proposed GRNMFHMDA model.

Main Methods:

• Constructed microbe and disease similarity networks using symptom-based disease similarity and Gaussian interaction profile kernel similarity.
• Implemented a preprocessing step to assign likelihood scores to unknown microbe-disease pairs.
• Utilized a graph regularized non-negative matrix factorization framework for simultaneous prediction of microbe-disease associations.

Main Results:

• Achieved reliable performance with Area Under the Curve (AUC) values of 0.8715 (global LOOCV) and 0.7898 (local LOOCV).
• Case studies on three human diseases showed that most top-ranked predicted microbes were validated by the HMDAD database or experimental literature.

Conclusions:

• The GRNMFHMDA model provides a reliable computational approach for predicting human microbe-disease associations.
• This model can significantly contribute to understanding the complex interplay between microbes and human diseases, aiding in future research and therapeutic strategies.