CRBPSA: CircRNA-RBP interaction sites identification using sequence structural attention model.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Circular RNAs (circRNAs) are crucial in gene regulation and disease prevention through interactions with RNA-binding proteins (RBPs).
• Existing identification algorithms often overlook secondary structure features, relying on limited sequence descriptors.
• Current methods face challenges with data sparsity and computational burden, impacting predictive accuracy.

Purpose of the Study:

• To develop an advanced computational tool for predicting circRNA-RBP binding sites.
• To effectively capture and utilize secondary structure features of circRNAs in prediction models.
• To improve the accuracy and efficiency of circRNA-RBP interaction prediction.

Main Methods:

• Developed CRBPSA, a novel architecture utilizing a sequence-structure attention mechanism.
• Generated base-pairing matrices with Gaussian-weighted secondary structures.
• Employed a Structure_Transformer for extracting spatial positional dependencies and base-pairing information.

Main Results:

• CRBPSA achieved an average AUC of 99.93% across 37 circRNA datasets (671,952 samples).
• The algorithm demonstrated superior performance compared to all existing prediction methods.
• Successfully identified circRNA-RBP binding sites through deep feature extraction.

Conclusions:

• CRBPSA is a lightweight and efficient tool for predicting circRNA-RBP binding sites.
• The tool accurately captures sequence structural features with minimal computational resources.
• Facilitates a deeper understanding of circRNA-protein interactions and their biological implications.