Identifying Potential miRNA-Disease Associations Based on an Improved Manifold Learning Framework

Yong Tang,Dong Yi,Yazhou Wu,Gaoming Li

doi:10.1109/access.2020.2972068

Abstract

Adequate evidence has shown that miRNA-disease interactions are strongly involved in the pathological processes of complex human diseases. However, it is commonly time-consuming and labor-intensive to utilize laboratory biological experiments to reveal unknown miRNA-disease pairs. Since the previously proposed calculation model has more or fewer deficiencies, we developed the semi-supervised method called Hessian Regularized Non-negative Matrix Factorization Method for miRNA-disease Association prediction (HRNMFMDA). This model introduced Hessian regularization into the NMF framework to preserve the local manifold information and increased an $l_{2,1}$ -norm penalty term to ensure the feature selection of the coding matrix and an approximate orthogonal constraint to obtain discriminative information. In the model performance evaluation, HRNMFMDA outperformed eight existing models, achieving AUC values of 0.9074, 0.8618, and 0.9044+/-0.0080 in the global leave-one-out cross-validation (LOOCV), local LOOCV and 5-fold cross-validation, respectively. In addition, we applied HRNMFMDA to several high-incidence human carcinomas via three kinds of case studies. Almost all predicted miRNAs were confirmed by external databases derived from experimental literature. Therefore, the conclusion can be drawn that HRNMFMDA is reliable for revealing uncovered miRNA-disease pairs.

Highlights

As a series of short non-coding RNAs of ∼22 nucleotides with a single strand [1], microRNAs are widely found in plants [2], animals [3], and certain viruses [4]
Studies on Alzheimer’s disease (AD) have revealed that lower expression levels of miR-15 and miR-107 in the hippocampus of AD brains lead to increased phosphorylation of tau protein, progressively inducing the occurrence of AD [16]
Experiments have demonstrated that miR-21, miR-494, and miR-1973 have encouraging potential as biomarkers for the diagnosis of classical Hodgkin’s lymphoma [17]

Summary

INTRODUCTION

As a series of short non-coding RNAs of ∼22 nucleotides with a single strand [1], microRNAs (miRNAs) are widely found in plants [2], animals [3], and certain viruses [4]. The method might cause bias to diseases associated with more miRNAs. Chen et al [42] developed an improved prediction model based on Laplacian Regularized Sparse Subspace Learning (LRSSLMDA). We presented an improved solution called Hessian Regularized Non-negative Matrix Factorization Method for miRNA-disease Association Prediction (HRNMFMDA) to further utilize information from negative samples. This model first introduced Hessian regularization into the NMF framework to exploit the local manifold structure by recovering the inherent local information from the training data. The results of the various verification experiments indicate that HRNMFMDA is useful for identifying candidate miRNA-disease pairs

MATERIALS AND METHODS

INTEGRATED SIMILARITIES FOR DISEASES AND MIRNAS

HESSIAN REGULARIZED NON-NEGATIVE MATRIX

PERFORMANCE VALIDATIONS

DISCUSSIONS