CRWS

CRWS predicts RNA-binding protein (RBP) binding sites on circular RNAs (circRNAs) using a stacked generalization ensemble deep learning model (CirRBP) to provide precise binding-site localization and motif identification.

Key Features:

• Multi-Source Data Integration: Integrates binding-site information from multiple databases to enrich the training dataset for CirRBP.
• Stacked Generalization Ensemble (CirRBP): Implements a stacked generalization ensemble that combines multiple deep learning frameworks.
• Weighted Prediction Aggregation: Aggregates component-model predictions through weighted averaging to improve predictive performance.
• Exact Binding Site Localization: Provides precise localization of RBP binding sites on circRNAs rather than only fragment-level probability values.
• Motif Discovery: Identifies the most widely distributed motifs within each RBP dataset to characterize common sequence and structural elements associated with RBP interactions.
• Sequence Input Flexibility: Accepts full-length circRNA sequences or fragments for prediction and localization.
• Binding-Site Visualization: Produces visual representations of predicted binding sites mapped onto input circRNA sequences.

Scientific Applications:

• circRNA–RBP interaction mapping: Maps RBP binding sites on circRNAs to support studies of circRNA–RBP interactions.
• Post-transcriptional regulation analysis: Facilitates investigation of circRNA roles in gene regulation and post-transcriptional regulatory mechanisms.
• Motif characterization: Enables identification of common sequence and structural motifs associated with specific RBPs on circRNAs.

Methodology:

CirRBP employs a stacked generalization ensemble of multiple deep learning frameworks trained on integrated multi-database binding-site information, aggregates predictions via weighted averaging, applies a binding-site localization algorithm, and performs motif discovery within each RBP dataset.