Comprehensive comparison of two types of algorithm for circRNA detection from short-read RNA-Seq

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are formed via back-splicing.
• Numerous software tools exist for circRNA detection, primarily using back-spliced junction reads.
• Emerging tools utilize k-mer tables or de Bruijn graphs instead of read mapping.

Purpose of the Study:

• To compare the performance of circRNA detection software based on different algorithms.
• To evaluate precision, sensitivity, and detection efficiency of eleven representative tools.
• To analyze circRNA detection with and without RNase R treatment across cell lines.

Main Methods:

• Comparative analysis of eleven circRNA detection tools.
• Utilized RNA-sequencing datasets from two cell lines, with and without RNase R treatment.
• Assessed performance using metrics like precision, sensitivity, AUC, F1 score, and detection efficiency.

Main Results:

• Tools based on reads mapping (CIRI2, KNIFE) demonstrated superior and balanced detection performance.
• Performance was consistent across different cell lines and RNase R treatment conditions.
• K-mer based tools showed comparatively lower detection efficiency.

Conclusions:

• Reads mapping-based tools offer more reliable circRNA detection compared to k-mer based approaches.
• The choice of circRNA detection tool impacts study outcomes.
• Further investigation into algorithmic differences is warranted for optimizing circRNA identification.