LPI-EnEDT: an ensemble framework with extra tree and decision tree classifiers for imbalanced lncRNA-protein...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Long noncoding RNAs (lncRNAs) play crucial roles in biological processes, and understanding their interactions with proteins is key to elucidating their functions.
• Experimental identification of lncRNA-protein interactions (LPIs) is resource-intensive.
• Existing computational methods struggle with the inherent imbalance in LPI data and often suffer from prediction bias due to reliance on single datasets.

Purpose of the Study:

• To develop a robust computational framework for predicting lncRNA-protein interactions (LPIs).
• To address the challenge of imbalanced LPI data classification.
• To improve the accuracy and reliability of LPI prediction by integrating diverse features and employing an ensemble learning approach.

Main Methods:

• Characterization of lncRNAs and proteins using Pyfeat and BioTriangle, followed by vector concatenation to represent lncRNA-protein pairs.
• Development of an ensemble framework (LPI-EnEDT) utilizing Extra Tree and Decision Tree classifiers for imbalanced LPI data classification.
• Validation across five distinct LPI datasets using rigorous cross-validation techniques.

Main Results:

• LPI-EnEDT demonstrated superior performance compared to four established LPI prediction methods (LPI-BLS, LPI-CatBoost, LPI-SKF, and PLIPCOM).
• Achieved high average AUC values (0.8480, 0.7078, 0.9066) and AUPR values (0.8175, 0.7265, 0.8882) across three cross-validation types on five datasets.
• Case analyses identified potential novel interactions, such as between HOTTIP and Q9Y6M1, and NRON and Q15717.

Conclusions:

• The proposed ensemble learning model effectively fuses diverse biological features for accurate lncRNA-protein interaction prediction.
• LPI-EnEDT provides a reliable method for classifying imbalanced LPI data and inferring novel interactions.
• This approach enhances our understanding of lncRNA functions and mechanisms by facilitating the identification of new interacting partners.