circDeep: deep learning approach for circular RNA classification from other long non-coding RNA

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Circular RNAs (circRNAs) are crucial regulators of microRNA activity and are implicated in diseases like cancer.
• Accurate detection of circRNAs is vital for understanding their biogenesis and function.
• Current methods for distinguishing circRNAs from other long non-coding RNAs (lncRNAs) have accuracy limitations.

Purpose of the Study:

• To develop a highly accurate and fast machine learning method for identifying circular RNAs.
• To improve the systematic annotation of circRNAs by addressing classification challenges.

Main Methods:

• An End-to-End deep learning framework named circDeep was developed.
• circDeep integrates RCM, ACNN-BLSTM sequence, and conservation descriptors.
• The framework utilizes high-level abstraction descriptors with shared representations across modalities.

Main Results:

• circDeep demonstrates superior performance compared to existing tools.
• The framework achieves a 12% increase in accuracy for circRNA identification.
• circDeep is significantly faster than current available tools.

Conclusions:

• circDeep offers a more accurate and efficient approach to circRNA classification.
• This advancement facilitates better understanding of circRNA roles in disease.
• The developed framework is publicly available for research use.