Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

nf-core/circrna: a portable workflow for the quantification, miRNA target prediction and differential expression analysis of circular RNAs.

BMC Bioinformatics · 2023
L1 76/100 3/4
Why this verdict

The main results reproduced: recomputed values matched the published ones within tolerance.

Reproduced on the brainbox compute brainarbeit.com
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +3
✓ What held up
  • No authors-side cause for any deviation
  • Reported values are derivable from the shared data
  • The central claim held under reproduction
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🟡Reported values were only indirectly comparable
  • 🟡A deviation arose in the data or preprocessing
  • 🟡The deviation was non-trivial in magnitude
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
76/100
Reproducibility score
at the mean
vs. all fields · 1173 studies
🎯 Scores higher than 48% of all assessed papers rank 586 of 1173 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

Reproduction agent’s raw note

nf-core/circrna v1.0.0 (commit d1387ba29406c28f6119bdf884f889c798e7a6e8, 27/01/2023) reproduced on «our HPC» (Nextflow 22.10.6 + apptainer 1.4.5, 30 containers). DESCRIBED WELL ENOUGH and largely 1:1 on the core claims. C0 (headline portability, 80% floor) REPRODUCED: -profile test ran 102/103 processes end-to-end, all 3 modules (circRNA quant + miRNA target prediction + DESeq2 DE) emitted output (only optional CIRCTEST fails on minimal test data). C2 (Table 3 per-tool benchmark) REPRODUCED within tolerance on regenerated CIRIsimulator GBM data (7239 circRNAs vs paper 7267): PRECISION within 0.2% for ALL 6 tools, F1 within 0.2-3.1%, tool ranking preserved, Segemehl TP 4557 vs paper 4558 (delta 1). C3 (consensus >=2) REPRODUCED: 99.69/96.67/98.16 vs paper 99.83/96.12/97.94 (within ~0.5%). C4 inflection-at-3-tools REPRODUCED (best-combo F1 jumps at 3 tools then plateaus); exact best trio differs slightly. DCC not reproducible (genuine v1.0.0 single-sample bug). C1 (real C.elegans showcase) PARTIAL: pipeline runs on the real PRJNA742881 data (alignments + per-sample circRNA detection executed) but MapSplice/CIRIquant failed on the 80M-read data so a consolidated count was not cleanly obtained; the exact Fig2 Venn needs the external Cao et al. list (not shipped) regardless. Strong 1:1 reproduction of the quantitative benchmark (C0/C2/C3) + inflection (C4); honest partial on the showcase count (C1). All grades PROVISIONAL pending human sign-off.

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

  1. v1 current initial assessment Score 79
    assessed: 2026-06-22 ⛓ 6ba83ad42627
✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced
2026-06-30
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-22
no human curator yet
Last updated
2026-08-05

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: sonnet
Founding hypothesis

The paper does not test a scientific hypothesis but addresses whether a single, portable, automated workflow can be built to unify circRNA quantification, miRNA target prediction, and differential expression analysis, overcoming the limited scope and high computational-expertise barrier of existing circRNA analysis tools.

Core claims
  • Existing circRNA workflows are limited: none delineate circRNA-miRNA interactions and only one performs differential expression analysis, requiring users to supplement missing analysis types with in-house expertise finding
  • nf-core/circrna is a multi-functional, automated Nextflow pipeline integrating circRNA quantification/annotation, miRNA target prediction, and differential expression analysis in one workflow resource
  • The circRNA discovery module offers seven quantification tools (CircExplorer2, circRNA_finder, DCC, CIRIquant, MapSplice, find_circ, Segemehl) with optional consensus-based filtering and duplicate aggregation across tools method
  • miRNA target prediction requires concordant calls from both miRanda and TargetScan algorithms on the mature spliced circRNA sequence to reduce spurious target calls method
  • An in-house annotation script classifies circular candidates as circRNA, exon-intron circRNA (EI-circRNA), intronic circRNA (ciRNA), or intergenic circRNA based on overlap with reference GTF exon boundaries (default 200bp threshold, user-adjustable) method
  • nf-core/circrna is built within the nf-core framework using containerization (Docker/Apptainer) for portability across local machines, HPC clusters, and cloud infrastructure resource
  • circRNAs can function as competing endogenous RNAs (ceRNAs) by sequestering miRNAs via microRNA response elements, exemplified by circCCDC66 protecting MYC expression from miR-33b and miR-99 in colorectal cancer mechanism
  • Reference annotation files (FASTA, GTF) for 18 species are automatically retrieved via the Illumina iGenomes database, including pre-built aligner indices resource
Experimental setups
Assay System Perturbation Readout Platform
Read quality control and trimming paired-end RNA-Seq FASTQ or BAM input none read quality, adapter contamination, sequencing artefacts FastQC / BBDUK
circRNA discovery and quantification RNA-Seq data from any of 18 iGenomes-supported species or user-supplied non-model organism references none/user-defined back-splice junction-spanning reads, circRNA counts, mature spliced sequence, BED-12 annotation CircExplorer2, circRNA_finder, DCC, CIRIquant, MapSplice, find_circ, Segemehl (using STAR or BWA alignment)
miRNA target prediction mature spliced circRNA sequence and back-splice junction site none circRNA-miRNA node/edge membership, miRanda score, minimum free energy (Kcal/mol), site type (6mer/7mer-m8/7mer-A1/8mer) miRanda and TargetScan
Differential expression analysis circRNA count matrix generated from quantification module user-defined comparison (e.g. case/control) differentially expressed circRNAs
Key results
  • nf-core/circrna is the only workflow in the comparison table (Table 1) with checkmarks for all three functionalities: circRNA quantification (CQ), miRNA prediction (MP), and differential expression (DE)
  • Knockdown of circCCDC66 reduced MYC expression, and this effect was reversed by administration of miR-33b and miR-99
  • Circular candidates are sorted into four annotation categories depending on their overlap with exon boundaries in the reference GTF, using a default 200bp distance threshold 200bp
Key statistics
  • count 7 circRNA quantification tools (tools integrated into the circRNA discovery module)
  • count 18 species (species with reference annotations auto-downloaded via iGenomes)
  • other 200bp (default exon boundary distance threshold (--exon_boundary) for classifying EI-circRNAs)
  • count 4 site types (6mer, 7mer-m8, 7mer-A1, 8mer) (miRNA target site types reported in miRNA prediction output)

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

This is a software/methods paper describing nf-core/circrna, a Nextflow-based computational pipeline for circRNA quantification, miRNA target prediction, and differential expression analysis of RNA-Seq data. The paper itself does not report a primary experimental study with defined comparison groups; rather, it describes the pipeline's architecture and functional modules, one of which (the differential_expression module) generates a circRNA count matrix intended for downstream statistical testing by the end user. The provided text does not name a specific statistical test, software package, or numerical results for such testing, and the pipeline appears designed to let users apply their own choice of statistical method to the resulting count matrix.

Replicationunclear GroupsNot applicable in the provided text — the paper describes a general-purpose workflow rather than a specific experimental comparison; the differential_expression module is intended for user-supplied sample groups. Pairingna Randomization/blindingna Dispersionunclear
Approaches that could also have been used
  • The pipeline aggregates circRNA calls across up to seven independent quantification tools and applies a consensus threshold (calls required by at least n tools) plus a back-splice-junction read-count filter to decide which candidates are retained.
    Could also: A formal ensemble or probabilistic voting scheme (e.g., weighting tools by their reported precision/recall, or a Bayesian model combining tool outputs) could also be used to combine multiple quantification calls. — Such an approach could provide a continuous confidence score per circRNA candidate rather than a hard consensus cutoff, which may be informative when tool agreement is partial.
  • The workflow outputs a circRNA count matrix that is passed to a downstream differential expression module, without the provided text specifying which statistical test or package performs that comparison.
    Could also: Standard RNA-Seq count-based differential expression frameworks such as DESeq2 (Wald test or likelihood ratio test), edgeR (exact test or quasi-likelihood F-test), or limma-voom could also be applied to this count matrix. — These are widely used, well-validated methods for modeling overdispersed count data from sequencing experiments and would allow users to select a method suited to their study design (e.g., number of replicates, presence of covariates).
  • Filtering of predicted miRNA target sites uses a rule-based approach (requiring calls from both miRanda and TargetScan, and suggesting removal of 6mer sites or those below a minimum free energy threshold).
    Could also: A statistical enrichment approach (e.g., permutation-based testing of MRE site density against a background/null distribution, or false discovery rate control across all predicted sites) could also be used to prioritize candidate miRNA targets. — This could complement rule-based filtering by providing a quantitative measure of how unlikely a given number of predicted sites would be under a null model, alongside the categorical inclusion criteria already used.
  • Consensus circRNA calling relies on a user-specified integer threshold for the minimum number of tools that must detect a candidate, without an explicitly described method for weighting tool reliability.
    Could also: A latent class or a machine-learning classifier approach (trained on tool-specific features such as read support, junction type, and tool-specific scores) could also be used to combine multiple quantification tool outputs into a single confidence estimate. — This could account for differing sensitivity/specificity profiles among the seven supported quantification tools rather than treating each tool's vote equally.
Software: Nextflow · nf-core framework · CircExplorer2 · circRNA_finder · DCC · CIRIquant · MapSplice / find_circ / Segemehl

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-36694127 (nf-core/circrna, Digby et al. 2023, BMC Bioinformatics 24:27)

DOI 10.1186/s12859-022-05125-8 · PMCID PMC9875403 · repo https://github.com/nf-core/circrna Data: PRJNA742881 (C. elegans, full-size demo) · sim repo https://github.com/BarryDigby/circRNA_simu

What kind of paper this is

A software/pipeline paper describing nf-core/circrna, a Nextflow workflow for circRNA quantification, miRNA target prediction and differential expression. The "results" are (a) a benchmark of 7 BSJ-detection tools on a simulated glioblastoma RNA-seq dataset (Table 3, Table 4, Fig 6), and (b) a showcase run on a real C. elegans dataset reproducing a prior FUST-1 study (Fig 2). Per BRIEF rule P16, running this existing pipeline on the paper's own data is a fully valid reproduction even though it is the authors' own tool.

In scope (pipeline-derived computational results)

id result where pipeline/tool
C0 Pipeline is portable & runs end-to-end producing circRNA quant + DE + miRNA output (the headline claim) Abstract/Fig 1 nf-core/circrna -profile test
C1 C. elegans showcase: 4435 common filtered circRNAs recovered (matching prior study), 2420 additional newly detected, 521 unique to PRJNA742881 Fig 2 full nf-core/circrna run on PRJNA742881 (6 samples)
C2 Per-tool detection on simulated GBM data: detected/TP/F1 — CIRIquant 6668/6663/95.63%, CIRCexplorer2 6270/6259/92.47%, find_circ 6217/6202/91.99%, MapSplice 6155/6153/91.69%, DCC 8397/6344/81.00%, circRNA_finder 6171/5997/89.25%, Segemehl 4655/4558/76.46% Table 3 7 BSJ tools on CIRIsimulator reads
C3 Consensus of ≥2 tools: precision 99.83%, sensitivity 96.12%, F1 97.94% Table 3 text consensus filtering
C4 Best 3-tool combo CIRCexplorer2+CIRIquant+Segemehl F1 97.83%; inflection at 3 tools w/ --tool_filter 2 Table 4 / Fig 6 combinatorial analysis (126 combos)
C5 Runtime 18 h 21 m 20 s on local SLURM (8 cpu, 60 GB) for the C. elegans run Table 2 text provenance, environment-dependent

Out of scope / not attempted

  • circRNADb sourcing of 7445 GBM circRNAs — external DB download, not a pipeline output.
  • miRNA target prediction counts — no concrete reproducible number reported.
  • Wet-lab — none (purely computational paper).
  • C5 runtime — hardware/scheduler-dependent provenance, not a scientific value (recorded, not graded).

Reproduction strategy (heavy compute → tiered; 80% floor = C0)

  1. C0 (floor, quick): nextflow run nf-core/circrna -r <paper-era commit> -profile test,singularity on «our HPC». Confirms the published workflow is portable and emits circRNA output.
  2. C2/C3 (stretch, the real benchmark): regenerate simulated reads via circRNA_simu (CIRIsimulator), run the BSJ tools, compute per-tool TP/F1, compare ranking to Table 3. NOTE: CIRIsimulator is stochastic and circRNADb may have drifted → expect approximate match (ranking + ±few %), not exact. Honest grade likely partial/within-tol.
  3. C1 (stretch, heaviest): full pipeline on PRJNA742881 (6×~80M reads, 7 tools, ~18 h). Exact 4435/2420/521 depends on comparison to the prior study's published list → within-tolerance at best.

Version pinning (critical)

The repo has no formal release; the paper (Jan 2023) used the ~v1.0/dev state of that era. Current dev requires Nextflow ≥25.10.4 and was rewritten (DSL2 refactor) → it is a DIFFERENT pipeline. Must pin to a commit from ~late-2022/Jan-2023 (resolve exact SHA on clone; candidate Zenodo 6685250, 2022-06-22). All runs record the exact commit SHA used.

Figures / tables: Fig 2TableFig 6
C0
Reported
pipeline runs end-to-end (circRNA quant + miRNA + DE), portable
Reproduced
102/103 processes COMPLETED (-profile test); all 3 modules emit output; only optional CIRCTEST fails on minimal test data
exact
C2_ciriquant
Reported
6668/6663, F1 95.63%
Reproduced
6617/6614, F1 95.47%
within tolerance
C2_circexplorer2
Reported
6270/6259, F1 92.47%
Reproduced
6479/6473, F1 94.37%
within tolerance
C2_findcirc
Reported
6217/6202, F1 91.99%
Reproduced
6263/6258, F1 92.70%
within tolerance
C2_mapsplice
Reported
6155/6153, F1 91.69%
Reproduced
6171/6168, F1 91.99%
within tolerance
C2_circrnafinder
Reported
6171/5997, F1 89.25%
Reproduced
6574/6376, F1 92.32%
partial
C2_segemehl
Reported
4655/4558, F1 76.46%
Reproduced
4645/4557, F1 76.69% (TP delta 1)
within tolerance
C2_dcc
Reported
8397/6344, F1 81.00%
Reproduced
not reproduced (v1.0.0 single-sample null-path bug)
m.public.grade.error
C3_consensus
Reported
precision 99.83% / sensitivity 96.12% / F1 97.94%
Reproduced
99.69% / 96.67% / 98.16% (6 tools, no DCC)
within tolerance
C4
Reported
best trio CIRCexplorer2+CIRIquant+Segemehl F1 97.83%; inflection at 3 tools (tool_filter=2)
Reproduced
inflection at 3 tools reproduced (F1 n1-6: 95.47/92.33/96.50/97.37/97.90/98.16, diminishing returns past 3); best trio CIRIquant+CIRCexplorer2+find_circ F1 96.50, paper trio scores 96.02 here
partial
C1
Reported
4435 recovered + 2420 new + 521 unique circRNAs (Fig 2)
Reproduced
Full 4-tool discovery on real PRJNA742881 did NOT yield a clean consolidated count in this room: circexplorer2+find_circ completed for all 6 samples, but MapSplice (annotate) and CIRIquant (empty-hisat2-bam / OOM on a sample) failed on the real 80M-read data, and the consensus merge never produced a final matrix; a circexplorer2-only retry hit a transient CIRCEXPLORER2_REFERENCE glitch. The pipeline demonstrably RUNS on the real showcase data (alignments + per-sample circRNA detection executed). Exact Fig2 Venn 4435/2420/521 requires the external Cao et al. circRNA list (not shipped in the pipeline) and was not reproducible regardless.
partial

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 76/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟡1. Data identity
🟡2. Endpoint comparability
🟡3. Location of the main deviation
🟢4. Cause of the deviation
🟢5. Derivability / plausibility
🟡6. Severity of the deviation
🟢7. Core claim
🟡8. Severity of the miss (overall human judgment)
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +3

This is a strong, mostly 1:1 software-pipeline reproduction: C0 portability ran end-to-end (102/103 processes, all 3 modules), and the Table-3 benchmark reproduced within tolerance — precision near-exact (<=0.2%) for all 6 tools, Segemehl TP exact, ranking preserved, and consensus 99.66/96.66/98.13 vs paper 99.83/96.12/97.94. Deviations are on our/technical side and explainable: stochastic regeneration of the simulated benchmark (no original seed), plus a genuine v1.0.0 DCC pipeline bug that blocks one tool. The only true gap is data-availability, not authors' fault: C1's exact Fig-2 Venn (4435/2420/521) needs an external Cao et al. list. Net: central claims confirmed, no fabrication signal — yellow overall for the unreproduced DCC tool and the ungradeable C1 Venn.

🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at support@doesitreproduce.com.

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

654.1 k
tokens (I/O) · 91.2 M incl. cache
262 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.