Prediction of Microbe-Disease Associations by Graph Regularized Non-Negative Matrix Factorization

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Microbes significantly influence human health and disease development.
• Understanding microbe-disease relationships is vital for pathogenesis insights, diagnosis, and prevention.
• Current knowledge of microbe-disease associations remains limited.

Purpose of the Study:

• To develop and validate an effective computational model for predicting microbe-disease associations.
• To enhance the understanding of the complex interplay between microbes and human diseases.
• To identify novel potential microbe-disease links for further investigation.

Main Methods:

• Proposed a novel method: Non-negative Matrix Factorization Microbe-Disease Associations (NMFMDA).
• Utilized Gaussian interaction profile kernel similarity for microbial and disease similarity calculations.
• Employed a graph-regularized non-negative matrix factorization model integrated with a logistic function for disease similarity regulation.
• Validated the method using fivefold cross-validation.

Main Results:

• Achieved a high Area Under the Receiver Operating Characteristic Curve (AUC) of 0.8891, outperforming existing state-of-the-art methods.
• Demonstrated robust performance in case studies involving asthma, inflammatory bowel disease, and colon cancer.
• Successfully identified potential microbe-disease associations with high accuracy.

Conclusions:

• NMFMDA is a powerful and effective computational tool for predicting microbe-disease associations.
• The method holds significant potential for advancing disease diagnosis and prevention strategies.
• Further research can leverage this model to explore the microbiome's role in various human ailments.