Predicting potential miRNA-disease associations by combining gradient boosting decision tree with logistic regression

Area of Science:

• Biomedical Informatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial in biological processes, and their dysregulation links to complex human diseases.
• Predicting miRNA-disease associations is vital but challenged by limited data and suboptimal existing methods.

Purpose of the Study:

• To develop a novel computational model for prioritizing miRNA candidates for various diseases.
• To enhance the accuracy of predicting potential miRNA-disease associations.

Main Methods:

• A hybrid Gradient Boosting Decision Tree with Logistic Regression (GBDT-LR) model was developed.
• K-means clustering was used for negative sample screening to balance datasets.
• GBDT was employed for feature extraction, followed by LR for final prediction.

Main Results:

• The GBDT-LR model achieved an average AUC of 0.9274 in 5-fold cross-validation.
• Case studies showed high confirmation rates for top predicted miRNAs associated with colon, gastric, and pancreatic cancers (90%, 94%, 88%).
• GBDT-LR outperformed three other state-of-the-art methods in prediction performance.

Conclusions:

• The GBDT-LR model offers a robust and accurate method for predicting miRNA-disease associations.
• This approach can significantly aid in identifying novel disease-related miRNAs and biomarkers.
• The model's performance suggests its utility in advancing research on complex diseases.