Is Circular RNA Detection Reliable Across Platforms?

Detection is reliable for abundant circRNAs and unreliable for the long tail. Three findings recur in benchmarking studies:

• • Concordance between computational callers on the same data is moderate; different pipelines return substantially different lists.
• • Concordance across wet-lab platforms (short-read sequencing, RNase R enrichment, arrays, long reads) is lower than expected, especially for low-abundance species.
• • Most disagreements involve junctions that are difficult to align, low-coverage loci, or sequences with genomic repeats.

Practical consequence: a circRNA reported by one platform and not another is not necessarily absent — but it also is not yet established.

For a circRNA you intend to study:

1. Divergent primers with Sanger sequencing of the junction product.
2. RNase R resistance, with a linear transcript as digestion control.
3. Genomic locus check — confirm no duplication or rearrangement generates the junction.
4. Absolute quantification with a spike-in, if abundance matters for the claim.
5. Orthogonal platform (long reads) for novel or low-abundance species.

Two of the first three is the usual minimum; the third is the one most often skipped and the one that catches the most consequential artefact.

• • Short-read sequencing vs arrays: arrays detect designed junctions only; absence is uninformative.
• • Short-read vs long-read: long reads confirm structure but have lower dynamic range; discordance is common at low abundance.
• • RNase R enrichment vs untreated libraries: enrichment changes relative abundance estimates, so ratios are not comparable across protocols.
• • Different callers on the same data: expect list overlap well below 100%; report the caller and thresholds.

Three cautions:

• • Normalize to total circular signal, not only to total reads, since RNase R efficiency varies by sample.
• • Use replicates; circRNA counts are sparse and single-sample comparisons are unstable.
• • Confirm the top candidates with targeted assays rather than reporting a list from one pipeline.

EDITGENE validates circRNA candidates with junction sequencing, RNase R resistance, genomic locus checks, spike-in absolute quantification, and orthologous long-read confirmation.

Tell us the candidate and the claim. Contact EDITGENE to request a project consultation or quote.

Platform concordance depends on abundance, library protocol, caller, and annotation; benchmarking results are protocol-specific and should not be generalized as fixed error rates.

Detection establishes the existence of a junction; it does not establish circular structure or function without orthogonal confirmation and functional assays.

• • Szabo L, Salzman J. Detecting circular RNAs: bioinformatic and experimental challenges. Nature Reviews Genetics. 2016;17:679–692.
• • Hansen TB, et al. Comparison of circular RNA prediction tools. Nucleic Acids Research. 2016;44:e58.
• • Kristensen LS, et al. The biogenesis, biology and characterization of circular RNAs. Nature Reviews Genetics. 2019;20:675–691.