Comparison of circular RNA prediction tools.

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Circular RNAs (circRNAs) are a class of non-coding RNAs whose abundance is increasingly recognized.
• circRNA identification relies on detecting back-splice junctions in sequencing reads.
• Numerous computational pipelines exist for circRNA prediction from RNA sequencing data.

Purpose of the Study:

• To compare the performance of five popular circRNA prediction algorithms: circRNA_finder, find_circ, CIRCexplorer, CIRI, and MapSplice.
• To evaluate the accuracy and reliability of these algorithms in identifying bona fide and false positive circRNAs.
• To assess the impact of algorithm choice on circRNA landscape prediction.

Main Methods:

• Utilized standard RNA sequencing datasets for comparative analysis.
• Applied five distinct circRNA prediction pipelines to the same datasets.
• Validated predicted circRNAs for bona fide status using RNase R resistance assays.

Main Results:

• Observed substantial discrepancies in circRNA predictions among the evaluated algorithms.
• Differences were particularly pronounced for highly expressed circRNAs and those originating from proximal splice sites.
• RNase R resistance assays highlighted varying levels of true positive circRNAs across different pipelines.

Conclusions:

• circRNA annotation requires careful consideration due to significant algorithmic variability.
• Combining outputs from multiple circRNA prediction tools is crucial for achieving reliable and robust circRNA landscape identification.
• Future research should focus on developing standardized and validated methods for circRNA discovery.