Potential miRNA-disease association prediction based on kernelized Bayesian matrix factorization.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial in human complex diseases.
• Understanding miRNA-disease associations aids in pathogenesis research and biomarker discovery.
• Experimental validation is time-consuming and costly, necessitating computational approaches.

Purpose of the Study:

• To develop a novel computational model, KBMFMDA, for predicting miRNA-disease associations.
• To enhance the understanding of molecular mechanisms underlying complex diseases.
• To identify potential diagnostic biomarkers for diseases.

Main Methods:

• Developed a Bayesian model (KBMFMDA) integrating kernel-based nonlinear dimensionality reduction, matrix factorization, and binary classification.
• Projected miRNAs and diseases into a unified subspace to estimate the association network.
• Employed rigorous cross-validation techniques (global and local leave-one-out, five-fold cross-validation).

Main Results:

• Achieved high performance with AUCs of 0.9132 (global LOO), 0.8708 (local LOO), and 0.9008±0.0044 (5-fold CV).
• Successfully applied KBMFMDA to three human cancers, identifying top potential disease-related miRNAs.
• Validated a majority of the top 50 predicted miRNAs through existing experimental reports.

Conclusions:

• KBMFMDA is a powerful and accurate computational tool for predicting miRNA-disease associations.
• The model facilitates the discovery of novel miRNA-disease links and potential biomarkers.
• This approach accelerates research into complex diseases and aids in developing diagnostic strategies.