ANMDA: anti-noise based computational model for predicting potential miRNA-disease associations

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are key regulators of gene function with established links to various diseases.
• Developing computational methods is crucial for uncovering novel miRNA-disease associations to guide further research.

Purpose of the Study:

• To address the impact of data noise on predicting miRNA-disease associations.
• To propose and validate an anti-noise algorithm (ANMDA) for enhanced prediction accuracy.

Main Methods:

• Feature construction based on miRNA and disease similarity.
• Negative sample generation using k-means clustering.
• LightGBM model training on data subsets with replacement.
• Ensemble prediction via a voting method.

Main Results:

• ANMDA achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.9373 ± 0.0005 in five-fold cross-validation.
• The algorithm outperformed several existing methods in predicting miRNA-disease associations.
• Case study analysis confirmed ANMDA's practical utility and novelty in inferring potential relationships.

Conclusions:

• Data noise significantly impacts the accuracy of miRNA-disease association predictions.
• ANMDA offers a robust approach to mitigate noise, improving prediction outcomes.
• Further enhancements are anticipated with advanced noise-handling techniques.