LRSSLMDA: Laplacian Regularized Sparse Subspace Learning for MiRNA-Disease Association prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Predicting microRNA (miRNA)-disease associations is crucial for identifying diagnostic biomarkers and therapeutic targets.
• Existing computational methods struggle with effectively integrating diverse biological datasets for reliable miRNA-disease association predictions.

Purpose of the Study:

• To develop a novel computational model, LRSSLMDA, for predicting miRNA-disease associations.
• To effectively integrate statistical and graph theoretical features of miRNAs and diseases into a common subspace for enhanced prediction accuracy.

Main Methods:

• Proposed Laplacian Regularized Sparse Subspace Learning for MiRNA-Disease Association prediction (LRSSLMDA).
• Utilized Laplacian regularization to preserve local data structures and L1-norm constraint for feature selection.
• Projected miRNA/disease features into a common subspace for dimensionality reduction and improved prediction.

Main Results:

• LRSSLMDA outperformed ten previous models in prediction accuracy, achieving AUCs of 0.9178 (global LOOCV) and 0.8418 (local LOOCV).
• Demonstrated high accuracy and stability with an average AUC of 0.9181+/-0.0004 in 5-fold cross-validation.
• Case studies validated top predictions for Colon Neoplasms, Lymphoma, Kidney Neoplasms, Esophageal Neoplasms, and Breast Neoplasms with high experimental evidence rates (88-98%).

Conclusions:

• LRSSLMDA is a robust and accurate computational tool for predicting miRNA-disease associations.
• The model's ability to integrate diverse data types and its high validation rates highlight its potential clinical significance.
• LRSSLMDA facilitates the discovery of novel miRNA biomarkers and therapeutic targets for various human diseases.