Integrating random walk and binary regression to identify novel miRNA-disease association

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) play crucial roles in human complex diseases.
• Advancements in data and methodologies enable computational prediction of miRNA-disease associations.

Purpose of the Study:

• To develop and validate a computational model for predicting miRNA-disease associations.
• To efficiently identify potential candidate miRNAs for various human complex diseases.

Main Methods:

• Proposed the Random Walk and Binary Regression-based MiRNA-Disease Association prediction (RWBRMDA) model.
• Extracted miRNA features using random walk with restart on an integrated miRNA similarity network.
• Employed binary logistic regression for association prediction.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.8076 in leave-one-out cross-validation.
• Case studies demonstrated high confirmation rates: 94% for Esophageal cancer, 90% for Prostate cancer, and 98% for Breast cancer.
• Lymphoma prediction showed 88% verification, confirming model robustness.

Conclusions:

• RWBRMDA can accelerate experimental investigations by providing testable, top-ranked miRNA candidates.
• The model offers a cost-effective approach to identifying disease-related miRNAs, reducing experimental effort.