iDRBP-ECHF: Identifying DNA- and RNA-binding proteins based on extensible cubic hybrid framework.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Proteins interacting with nucleic acids are vital for regulating biological processes.
• Accurate identification of nucleic acid-binding proteins (NABPs) is significant for understanding cellular functions.
• Existing computational methods for NABP identification often use imbalanced datasets and traditional machine learning, limiting performance.

Purpose of the Study:

• To develop a novel sequence-based computational method for predicting DNA-binding proteins (DBPs) and RNA-binding proteins (RBPs).
• To address limitations in existing methods by constructing a more realistic benchmark dataset and incorporating advanced algorithms.

Main Methods:

• Proposed iDRBP-ECHF, a sequence-based method for DBP and RBP prediction.
• Constructed a benchmark dataset reflecting real-world sample proportions and applied down-sampling for balanced training sets.
• Integrated deep learning algorithms to extract high-level feature representations.

Main Results:

• The iDRBP-ECHF method achieved state-of-the-art performance on independent test datasets.
• Demonstrated superior prediction accuracy compared to existing sequence-based DBP and RBP identification methods.
• Developed a user-friendly webserver (http://bliulab.net/iDRBP-ECHF) for the iDRBP-ECHF tool.

Conclusions:

• The iDRBP-ECHF method offers a significant advancement in predicting DNA- and RNA-binding proteins.
• The use of balanced datasets and deep learning enhances prediction accuracy and reliability.
• The publicly available webserver facilitates broader application and research in NABP identification.