Identifying and Exploiting Potential miRNA-Disease Associations With Neighborhood Regularized Logistic Matrix...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNAs (miRNAs) are crucial in biological research, with many studies linking them to diseases.
• Experimental validation of miRNA-disease associations is costly and time-consuming.
• Computational methods are essential for efficient prediction of miRNA-disease associations.

Purpose of the Study:

• To develop a novel computational method, NRLMFMDA, for predicting miRNA-disease associations.
• To integrate multiple data sources for enhanced prediction accuracy.
• To validate the proposed method's effectiveness and practicality.

Main Methods:

• Proposed a neighborhood regularized logistic matrix factorization method (NRLMFMDA).
• Integrated miRNA functional similarity, disease semantic similarity, and Gaussian interaction profile kernel similarity.
• Incorporated neighborhood information and weighted known associations to improve prediction.

Main Results:

• Achieved high AUC values in global (0.9068) and local (0.8239) leave-one-out cross-validation.
• Demonstrated an average AUC of 0.8976 ± 0.0034 in 5-fold cross-validation.
• Case studies showed high validation rates (78-80%) for predicted miRNA-disease associations in specific cancers and lymphomas.

Conclusions:

• NRLMFMDA significantly outperforms previous models in predicting miRNA-disease associations.
• The method is effective and practical for identifying potential miRNA-disease links.
• This approach can accelerate biological research by reducing experimental burden.