Predict potential miRNA-disease associations based on bounded nuclear norm regularization

Area of Science:

• Computational Biology and Bioinformatics
• Genomics and Molecular Biology
• Disease Association Studies

Background:

• Abnormal microRNA (miRNA) expression is linked to complex human diseases.
• Experimental validation of miRNA-disease associations is costly and time-consuming.
• Accurate computational prediction of miRNA-disease associations is crucial.

Purpose of the Study:

• To develop an accurate computational model for predicting potential miRNA-disease associations.
• To address limitations in existing prediction methods, aiming for improved accuracy.
• To offer a cost-effective and efficient alternative to experimental validation.

Main Methods:

• Proposed a novel matrix completion model with bounded nuclear norm regularization (BNNRMDA).
• Constructed a heterogeneous miRNA-disease network integrating miRNA/disease self-similarity and known associations.
• Employed alternating direction method for solving the matrix completion problem with error tolerance and value boundary.

Main Results:

• BNNRMDA demonstrated superior performance compared to four state-of-the-art methods in cross-validation.
• Achieved best performance in both five-fold and leave-one-out cross-validation.
• Case studies showed 47 of the top 50 predictions were validated in the HMDD database.

Conclusions:

• BNNRMDA effectively leverages integrated information for accurate miRNA-disease association prediction.
• The model is robust to noise and offers significant improvements over existing methods.
• BNNRMDA provides a valuable tool for accelerating the discovery of disease-related miRNAs.