Integrating multi-encoding sequence features via stacking ensemble learning for RNA m5C site prediction.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA 5-methylcytosine (m5C) is a crucial epitranscriptomic modification impacting RNA stability and translation.
• Current computational methods struggle with accurate m5C site identification due to limited sequence representation and feature integration.

Purpose of the Study:

• To develop a comprehensive machine learning framework for enhanced RNA m5C site prediction.
• To overcome limitations of existing methods by integrating diverse sequence encoding schemes.

Main Methods:

• Integration of six sequence encoding schemes: ENAC, TNC, CKSNAP, PseEIIP, one-hot encoding, and NCP.
• Utilizing a stacking ensemble strategy to combine outputs from optimal base classifiers.
• Employing 5-fold and 3-fold cross-validation for robust model training and validation.

Main Results:

• Achieved high performance with 75.5% accuracy, 0.51 MCC, and 0.82 PR-AUC.
• Demonstrated improved robustness and cross-dataset generalization capabilities.
• The fusion-based ensemble framework proved effective for RNA m5C site prediction.

Conclusions:

• The proposed machine learning framework offers an effective and reliable solution for predicting RNA m5C sites.
• Integration of multiple sequence features and ensemble learning enhances prediction accuracy and generalization.
• This work advances epitranscriptomic research by providing a powerful tool for m5C site identification.