Evaluation of methods to detect circular RNAs from single-end RNA-sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Circular RNAs (circRNAs) are RNA molecules with diverse biological functions and serve as potential disease biomarkers.
• Current circRNA detection methods often rely on paired-end (PE) RNA sequencing (RNA-Seq) data.
• The performance of these methods on single-end (SE) RNA-Seq data, common in datasets like CLIP-Seq, is not well understood.

Purpose of the Study:

• To systematically evaluate the performance of circRNA detection methods using SE RNA-Seq data.
• To compare RNA-based and CLIP-Seq based methods for circRNA identification.
• To investigate factors influencing detection accuracy, including read length and sequencing depth.

Main Methods:

• Evaluated six RNA-based and two CLIP-Seq based methods for circRNA detection from SE RNA-Seq data.
• Assessed method performance using precision, sensitivity, F1 score, and true discovery rate.
• Utilized simulated SE RNA-Seq datasets and real datasets including RNA-Seq and CLIP-Seq samples.

Main Results:

• Increasing read length or sequencing depth generally improves sensitivity for most methods.
• False positive rates significantly impact the precision of all evaluated methods.
• CLIP-Seq based methods outperform RNA-Seq based methods on short-read SE data, while RNA-based methods are better for long-read data.

Conclusions:

• No single circRNA detection method is optimal for all SE RNA-Seq data types.
• CLIP-Seq based methods are recommended for short-read SE RNA-Seq data, especially when combined.
• This evaluation aids researchers in selecting appropriate strategies for circRNA analysis from SE RNA-Seq data.