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

Many biological experimental studies have confirmed that microRNAs (miRNAs) play a significant role in human complex diseases. Exploring miRNA-disease associations could be conducive to understanding disease pathogenesis at the molecular level and developing disease diagnostic biomarkers. However, s...

Full description

Saved in:
Bibliographic Details
Published in:Genomics (San Diego, Calif.) Vol. 112; no. 1; pp. 809 - 819
Main Authors: Chen, Xing, Li, Shao-Xin, Yin, Jun, Wang, Chun-Chun
Format: Journal Article
Language:English
Published: United States Elsevier Inc 01.01.2020
Subjects:
Association prediction
Bayesian algorithm
Bayesian theory
biomarkers
case studies
Conjugate probabilistic model
Disease
humans
Matrix factorization
microRNA
neoplasms
pathogenesis
prediction
microRNA
Matrix factorization
Bayesian algorithm
Disease
Conjugate probabilistic model
Association prediction
ISSN:0888-7543, 1089-8646, 1089-8646
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Many biological experimental studies have confirmed that microRNAs (miRNAs) play a significant role in human complex diseases. Exploring miRNA-disease associations could be conducive to understanding disease pathogenesis at the molecular level and developing disease diagnostic biomarkers. However, since conducting traditional experiments is a costly and time-consuming way, plenty of computational models have been proposed to predict miRNA-disease associations. In this study, we presented a neoteric Bayesian model (KBMFMDA) that combines kernel-based nonlinear dimensionality reduction, matrix factorization and binary classification. The main idea of KBMFMDA is to project miRNAs and diseases into a unified subspace and estimate the association network in that subspace. KBMFMDA obtained the AUCs of 0.9132, 0.8708, 0.9008±0.0044 in global and local leave-one-out and five-fold cross validation. Moreover, KBMFMDA was applied to three important human cancers in three different kinds of case studies and most of the top 50 potential disease-related miRNAs were confirmed by many experimental reports. •The computational model made use of Bayesian inference and dimensionality reduction.•The AUCs of LOOCV and 5-fold cross validation were significantly better than many previous computational models.•Three case studies for important human diseases were performed.•KBMFMDA could be a reliable method for miRNA-disease association prediction.
Bibliography:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Undefined-1
ObjectType-Feature-3
content type line 23
ObjectType-Article-1
ObjectType-Feature-2
ISSN:0888-7543
1089-8646
1089-8646
DOI:10.1016/j.ygeno.2019.05.021