MLFLHMDA: predicting human microbe-disease association based on multi-view latent feature learning.

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Microorganisms are vital for human health, with imbalances linked to disease.
• Identifying microbe-disease associations aids in understanding disease pathogenesis.
• Traditional experimental methods for association discovery are costly and time-consuming.

Purpose of the Study:

• To develop a novel computational model, MLFLHMDA, for predicting human microbe-disease associations.
• To improve the accuracy and efficiency of identifying potential microbe-disease links.
• To provide a valuable tool for research in microbial roles in human health.

Main Methods:

• Utilized a Multi-View Latent Feature Learning approach.
• Computed Gaussian interaction profile kernel similarity and applied weighting K nearest known neighbors (WKNKN) for preprocessing.
• Extracted latent features, projected them into a common subspace, and incorporated graph regularization and L-norms for interpretability and sparsity.

Main Results:

• Achieved high performance with AUC values of 0.9165 (global leave-one-out) and 0.8942+/-0.0041 (5-fold cross-validation).
• Demonstrated superior predictive power compared to existing methods through case studies.
• The model's source code and data are publicly available.

Conclusions:

• MLFLHMDA effectively predicts human microbe-disease associations.
• The model offers a cost-effective and accurate alternative to traditional experimental methods.
• This work contributes to a deeper understanding of the human microbiome's role in health and disease.