Identifying the sequence specificities of circRNA-binding proteins based on a capsule network architecture

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Circular RNAs (circRNAs) are crucial regulators in biological processes and diseases.
• Understanding circRNA interactions with RNA-binding proteins (RBPs) is key to elucidating their functions.

Purpose of the Study:

• To develop a novel computational method for identifying sequence specificities in circRNA-RBP binding.
• To enhance the understanding of molecular mechanisms underlying circRNA-RBP interactions.

Main Methods:

• A modified capsule network, termed circRB, was developed.
• Convolution operations extract circRNA sequence features.
• Dynamic routing algorithms within the capsule network discriminate binding sites.

Main Results:

• The circRB method demonstrated superior performance compared to existing computational approaches.
• Sequence motifs were identified on circRNA-RBP binding datasets, with some overlapping known human RNA motifs.
• Predicted binding sites on full-length circRNAs can aid ongoing research.

Conclusions:

• The circRB classification framework, based on capsule networks, effectively addresses the limited research on circRNA-RBP sequence specificities.
• circRB achieves higher prediction accuracy, offering a valuable tool for the field.