BiLSTM- and CNN-Based m6A Modification Prediction Model for circRNAs.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• N6-methyladenosine (m6A) methylation is a crucial RNA modification impacting gene expression and cellular processes.
• While m6A prediction models exist for messenger RNA (mRNA), a gap persists for circular RNAs (circRNAs).
• Accurate prediction of m6A sites on circRNAs is vital for understanding their regulatory roles and involvement in diseases.

Purpose of the Study:

• To develop a novel computational model for precise prediction of m6A methylation sites in circRNAs.
• To leverage the strengths of Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks for enhanced prediction accuracy.

Main Methods:

• A hybrid CNN-BiLSTM model was designed, integrating an attention mechanism.
• The model was trained and validated using m6A methylation data from HEK293 cells.
• Feature extraction capabilities of CNNs and long-range dependency handling of BiLSTMs were synergistically employed.

Main Results:

• The developed model achieved over 78% prediction accuracy on independent datasets.
• The attention mechanism effectively highlighted critical biological information for circRNA m6A methylation.
• The model demonstrated superior performance in identifying m6A sites within circRNA sequences.

Conclusions:

• The novel CNN-BiLSTM hybrid model provides a valuable tool for accurate circRNA m6A site prediction.
• This advancement enhances the understanding of circRNA function and regulation in biological systems.
• The findings lay a foundation for future biomedical applications targeting circRNA methylation.