Attention-Based Deep Multiple-Instance Learning for Classifying Circular RNA and Other Long Non-Coding RNA.

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Circular RNAs (circRNAs) are a class of long non-coding RNAs (lncRNAs) involved in crucial biological processes and transcriptional regulation.
• Accurate identification of circRNAs from other lncRNAs is essential for advancing research in RNA biology.

Purpose of the Study:

• To develop a novel deep learning model for the accurate identification of circRNAs.
• To identify key sequence features and motifs associated with circRNA formation using an attention mechanism.

Main Methods:

• Designed an attention-based multi-instance learning (MIL) network architecture.
• Utilized raw RNA sequences as input to learn sparse sequence features.
• Validated the attention mechanism's effectiveness using a handwritten digit dataset.
• Performed motif enrichment analysis on identified key sequence loci.

Main Results:

• The proposed MIL network outperformed existing state-of-the-art models in circRNA identification.
• The attention mechanism successfully highlighted key sequence loci critical for circRNA recognition.
• Motif enrichment analysis revealed significant motifs involved in circRNA formation.

Conclusions:

• The developed deep learning architecture is effective for learning gene sequences with sparse features.
• The model demonstrates strong representational capabilities in identifying key loci for circRNA identification.
• This work provides a valuable tool for circRNA research and understanding RNA regulatory mechanisms.