iDRBP-EL

iDRBP-EL predicts DNA- and RNA-binding proteins from protein primary sequences using hierarchical ensemble learning to identify proteins that bind to DNA and RNA.

Key Features:

• Input: Uses protein primary sequences as the sole biological input.
• Prediction targets: Performs simultaneous prediction of DNA-binding proteins (DBPs) and RNA-binding proteins (RBPs) as a multi-label task.
• Algorithmic framework: Employs hierarchical ensemble learning to integrate information across multiple levels.
• Feature integration: Fuses diverse feature sets and different data sources within the ensemble framework.
• Model composition: Combines various machine learning algorithms into a cohesive multi-label model.
• Cross-prediction mitigation: Specifically addresses the cross-prediction problem between DBPs and RBPs through information fusion.
• Performance evaluation: Includes ablation experiments demonstrating that feature fusion improves prediction performance.
• Validation: Assessed using independent datasets and reported to outperform existing methods for DBP and RBP prediction.

Scientific Applications:

• Protein–nucleic acid interaction studies: Identification of proteins that bind DNA and/or RNA to inform mechanistic studies of binding specificity.
• Genomics and transcriptomics: Annotation of protein binding roles in genomic and transcriptomic datasets.
• Molecular biology: Prioritization of candidate DBPs and RBPs for experimental characterization in molecular biology research.

Methodology:

Uses protein primary sequences; constructs a hierarchical ensemble that fuses diverse feature sets, various machine learning algorithms, and different data sources into a multi-label model; conducts ablation experiments to assess the contribution of fused information and validates performance on independent datasets.