RNMFMDA: A Microbe-Disease Association Identification Method Based on Reliable Negative Sample Selection and Logistic...

Area of Science:

• Microbiology
• Computational Biology
• Genomics

Background:

• Abnormal microbial levels are linked to complex diseases.
• Identifying microbe-disease associations (MDAs) is crucial for understanding disease mechanisms.
• Experimental MDA identification is resource-intensive and slow.

Purpose of the Study:

• To develop a novel computational model, RNMFMDA, for predicting microbe-disease associations (MDAs).
• To improve the efficiency and accuracy of MDA identification compared to existing methods.

Main Methods:

• Developed RNMFMDA, a computational model integrating Positive-Unlabeled (PU) learning and random walk with restart.
• Employed Logistic Matrix Factorization with Neighborhood Regularization (LMFNR) for association probability calculation.
• Validated RNMFMDA against five state-of-the-art methods using five-fold cross-validation.

Main Results:

• RNMFMDA achieved superior performance with AUCs of 0.6332, 0.8669, and 0.9081 in three cross-validation scenarios.
• The model significantly outperformed existing MDA prediction methods.
• High association scores for predicted microbe-disease pairs warrant further experimental investigation.

Conclusions:

• RNMFMDA offers a highly effective computational approach for identifying microbe-disease associations.
• The model's success is attributed to robust negative sample selection, the LMFNR prediction engine, and integrated biological data.
• RNMFMDA provides a valuable tool for accelerating research into the role of microbes in complex diseases.