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

Chemical reversible crosslinking enables measurement of RNA 3D distances and alternative conformations in cells.

Nat Commun · 2022
L1 84/100 3/4
Why this verdict

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

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • No authors-side cause for any deviation
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
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The deviation was non-trivial in magnitude
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
84/100
Reproducibility score
0.6 SD above mean
vs. all fields · 1173 studies
🎯 Scores higher than 63% of all assessed papers rank 392 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

PARTIAL (well-reproduced core). SHARC-seq / CRSSANT pipeline, COMPUTE RAN on «our HPC» SLURM compute nodes. (1) ALL 4 shipped deterministic CRSSANT tests reproduced: gaptypes classification (10093->9179/339/1/1/281/0) EXACT; gapfilter splice/short-gap (352->291) byte-identical SAM EXACT; gapmcluster human-7SK TG assembly (275 alignments) byte-identical EXACT; crssant DG clustering (24 DGs) byte-identical in 3/5 runs, within-tol (clique-iteration non-determinism). 3/4 byte-identical -> strong evidence the read-processing + clustering core is faithful. (2) Headline biological claim '17 DGs on 7SK' (Supp Fig 10f) reproduced END-TO-END FROM RAW FASTQ (SRR13797237): STAR (exact README params) -> gaptypes -> gapfilter -> crssant on the 939 N-gapped reads with both arms inside 7SK -> 24-25 RN7SK DGs (5 runs); the top-17 by read support (>=5 reads) form a well-supported core consistent with the reported 17, extra 7-8 are low-support (2-5 reads) removed by a coverage/confidence threshold -> graded partial (same regime; not an exact-17 claim). (3) Dataset GSE167812: 12/12 samples present, profiled; deposit = processed read-arm tables (HeLa _anno.txt + in-vitro 1HR2 bedpe), NOT final per-RNA DG BEDPE; deepest HeLa lib dominated by CVA21 virus + rRNA. (4) Two latent bugs in shipped crssant.py line2info() documented (split '\n' vs '\t'; undefined genesdict) - dead SA-chimeric path never run by the authors' tests; NOT a fabrication signal. NOT reproduced: per-library exact %gapped for all 10 libraries (S1, partial). OUT OF SCOPE (not attempted): wet-lab crosslinking-efficiency assays, 3D distance/structural modeling (not pipeline-derived).

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 50
    assessed: 2026-06-19 ⛓ 06e39fa3fb0b
✎ 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
2026-06-25
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-19
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 authors test whether reversible bifunctional 2′-hydroxyl acylation crosslinkers of defined length (SHARC), combined with exonuclease trimming and proximity ligation sequencing, can measure spatial distances between nucleotides in RNA and thereby capture 3D tertiary structures and alternative conformations of RNA in living cells.

Core claims
  • SHARC uses chemical crosslinkers of defined lengths to measure distances between nucleotides in cellular RNA method
  • SHARC-exo (crosslinking + exo trimming + proximity ligation + sequencing) enables transcriptome-wide RNA tertiary structure contact maps at high accuracy and precision method
  • SHARC data provide constraints that improve Rosetta-based RNA 3D structure modeling to near-nanometer resolution finding
  • Integrating SHARC-exo with other crosslinking-based methods reveals compact folding of the 7SK RNA, a regulator of transcriptional elongation finding
  • Aromatic dicarboxylic acid-derived crosslinkers (e.g., dipicolinic acid) achieve near-quantitative (97-99%) RNA crosslinking efficiency finding
  • SHARC crosslinks can be selectively reversed under mild alkaline conditions without causing RNA phosphodiester chain damage finding
  • Exonuclease (RNase R) trimming pinpoints crosslink sites at near-nucleotide resolution by stalling ~5 nt from the crosslinked base method
  • Proteinase K/TNA extraction removes proteins prior to crosslink detection, indicating detected interactions are RNA-RNA rather than protein-mediated finding
Experimental setups
Assay System Perturbation Readout Platform
Crosslinking efficiency assay (polyacrylamide gel electrophoresis) model self-complementary RNA 1 duplex, in vitro treatment with 8 activated dicarboxylic acid crosslinkers % crosslinked RNA urea-denatured TBE PAGE gel
1H NMR hydrolysis kinetics dipicolinic acid imidazolide (DPI), in vitro pH 7.4 buffer, room temperature, time course hydrolysis half-life / rate constant NMR
Crosslink reversal / RNA stability assay model RNA 1 duplex crosslinked with DPI, in vitro; model dinucleotide ApA and compound 2 (DPI methyl ester) alkaline conditions (Borate buffer pH 10.0-11.0, 37°C) reversal efficiency and RNA degradation 1H NMR and urea-denatured TBE PAGE gel
SHARC-exo (crosslinking, RNase III digestion, DD2D gel isolation, RNase R exo trimming, proximity ligation, RT, high-throughput sequencing) HEK293T cells (ribosome, 7SL, RNase P, spliceosome, 7SK RNA) DPI crosslinking (5, 12.5, 25 mM) with varying RNase R trimming times gapped-read fraction, crosslinked fragment recovery %, inter-nucleotide spatial distances high-throughput sequencing
PARIS psoralen crosslinking with exo trimming validation cells (28S rRNA) psoralen crosslinking followed by RNase R trimming positional enrichment of uridine at trimmed 3′ ends high-throughput sequencing
icSHAPE reactivity comparison human ribosome, HEK293T cells none (comparison to existing icSHAPE dataset) icSHAPE signal along gapped-read arms
Proteinase K digestion / TNA RNA extraction HEK293T cells proteinase K treatment protein removal efficiency
Key results
  • Aromatic dicarboxylic acid crosslinkers (terephthalic, isocinchomeronic, dipicolinic acids) showed 97-99% crosslinking efficiency, versus 1-24% for oxalic/succinic acid 97-99%
  • DPI hydrolysis half-life at pH 7.4 was ~5 min, faster than the related SHAPE reagent NAI (~30 min half-life) ~6-fold faster
  • Compound 2 (DPI methyl ester) hydrolyzed far faster than the ApA phosphodiester bond, allowing selective crosslink reversal ~1000-fold difference in rate constant (3.5×10⁻⁴ s⁻¹ vs <4.0×10⁻⁷ s⁻¹)
  • SHARC crosslinked RNA duplex was nearly fully reversed after alkaline treatment without apparent degradation
  • Trimmed ribosome samples showed enrichment of single-stranded nucleotides peaking at the 5th nucleotide from the 3′ end ~1.3-fold over non-trimmed
  • icSHAPE reactivity signal was enriched near the crosslink site in trimmed samples ~3.7-fold
  • SHARC-exo measured ribosome inter-nucleotide distances close to the physical crosslinker length, with distances significantly narrower than shuffled-read controls mode ~8 Å; 51% within 20 Å (unrefined), 31-49% within 20-40 Å (refined)
  • Distances constrained by secondary structure (dsRNA) were predominantly short, while core and expansion-segment tertiary contacts showed progressively broader distance distributions 96.2% (dsRNA) vs 58.2% (core) vs 33.9% (ES) within 20 Å
Key statistics
  • fold_change 97-99% crosslinking efficiency (aromatic reagents) vs 1-24% (oxalic/succinic) (crosslinking efficiency of SHARC reagents on model RNA 1 duplex)
  • other DPI hydrolysis half-life ~5 min at pH 7.4 (DPI reaction kinetics measured by NMR)
  • other rate constant: ApA <4.0×10⁻⁷ s⁻¹; compound 2 3.5×10⁻4 s⁻¹ (R²=0.99); ~1000-fold difference (selective crosslink reversal vs phosphodiester stability)
  • pvalue p < 10⁻³⁰⁰ (Wilcoxon rank-sum test) (SHARC-exo distance distribution vs randomly shuffled reads, ribosome)
  • count 51% of minimal distances within 20 Å, mode ~8 Å (ribosome cryo-EM distance validation, all gapped reads)
  • count 31% within 20 Å and 49% within 40 Å (ribosome distances restricted to 5th±2 nt positions)
  • count dsRNA 96.2%, core 58.2%, expansion segments 33.9% of distances within 20 Å (distances by structural category (secondary vs tertiary contacts))
  • count 1.01%, 1.31%, 1.89% RNA fragments recovered as crosslinked at 5, 12.5, 25 mM DPI; 3.3-14.5% gapped reads (SHARC-exo crosslinking and library statistics in HEK293T cells)

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 methods paper presents SHARC, a chemical crosslinking-sequencing approach for measuring RNA inter-nucleotide distances in cells. Quantitative benchmarking relies primarily on Wilcoxon rank-sum tests comparing observed distance distributions to shuffled null distributions, with crosslinking efficiencies characterized as mean ± SD from technical replicates. Hydrolysis kinetics are summarized by rate constants with goodness-of-fit (R²). Most structural results are reported descriptively as cumulative distance percentages and modes rather than through formal inferential tests.

Replicationmixed Sample sizen=3 technical replicates for in vitro crosslinking efficiency (Fig. 1e); two biological replicates for SHARC-exo in cells (Fig. 2c,d); no formal power calculation mentioned GroupsCrosslinked reads vs. randomly shuffled reads (distance distributions); different DPI concentrations (5, 12.5, 25 mM); trimmed vs. non-trimmed SHARC; dsRNA vs. tertiary contacts (core vs. expansion segments) in ribosome Pairingunclear Randomization/blindingnot stated DispersionSD Exact p-valuesno Effect sizesno Confidence intervalsno Multiplicity correctionno
Statistical tests used
Test Applied to n Assumptions
Wilcoxon rank-sum (WRS) test Comparison of minimal inter-nucleotide distance distributions between SHARC-exo gapped reads and randomly shuffled reads in the human 28S rRNA (Fig. 2e and Fig. 2g) All gapped reads mapped to the human ribosome; exact count not stated in provided text not stated
Exponential decay curve fitting (rate constant estimation) Hydrolysis kinetics of model SHARC compound 2 and ApA dinucleotide (Fig. 1h); R²=0.99 reported for compound 2 not stated
Approaches that could also have been used
  • p-values from Wilcoxon rank-sum tests were reported only as bounds (p < 10^-300) without accompanying effect sizes
    Could also: A rank-biserial correlation or median difference with 95% CI could also be reported alongside the p-value — With very large n (many sequencing reads), even trivially small differences produce near-zero p-values; effect sizes quantify the magnitude of the difference independently of sample size and support comparison across experiments or datasets
  • Distance distributions were compared to a shuffled null using the Wilcoxon rank-sum test, which is sensitive to location shift
    Could also: A Kolmogorov-Smirnov (KS) test or a permutation test on a chosen summary statistic could also be used — KS and permutation tests are sensitive to differences in the full distributional shape—including spread and tail behavior—which may be particularly informative given the explicitly described long-tailed distance distributions and heterogeneous RNA conformations
  • Crosslinking efficiency experiments used n=3 technical replicates (same RNA preparation, repeated measurements)
    Could also: Independent biological replicates (separate RNA preparations and crosslinking reactions) could also be used — Biological replicates capture preparation-to-preparation variability and support broader inference about the method's reproducibility across experimental batches, complementing technical replicates that primarily reflect measurement precision
  • Dispersion was reported as SD from a small number of replicates (n=3 technical; n=2 biological)
    Could also: 95% confidence intervals could also be reported alongside or instead of SD — CIs directly express uncertainty in the estimated mean and are often recommended for small-n experiments because they scale with n, whereas SD reflects only the spread of observations and stays roughly constant as n grows
  • Multiple comparisons were performed across DPI concentrations, trimming conditions, and RNA structural categories without a stated multiplicity correction
    Could also: A Benjamini-Hochberg FDR correction or Bonferroni correction could also be applied across the family of comparisons — Applying a correction makes explicit which findings remain significant after accounting for the number of tests and clarifies the family-wise error rate, which is a standard consideration when many comparisons are reported in the same study
  • Structural subpopulations (dsRNA, core tertiary, expansion segment) were characterized descriptively by the percentage of reads falling within fixed distance thresholds
    Could also: Mixture modeling (e.g., a mixture of log-normal or gamma components fit to the distance distributions) could also formally decompose reads into structural subpopulations — Mixture models can simultaneously estimate the proportion and characteristic distance parameters of each conformational state, providing a quantitative complement to the threshold-based summaries used and potentially better capturing the heterogeneous conformations the paper emphasizes
Software: Rosetta

What was reproduced

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

Scope — pmid-35177610 (SHARC-seq / CRSSANT)

Paper: Van Damme et al. 2022, Nat Commun 13:911. "Chemical reversible crosslinking enables measurement of RNA 3D distances and alternative conformations in cells." (method = SHARC-seq)

  • DOI 10.1038/s41467-022-28602-3 · PMCID PMC8854666
  • Analysis code: https://github.com/zhipenglu/CRSSANT (third-party/lab tool, P16 OK)
  • Data: GEO GSE167812 (SRA SRP308411); processed = GSE167812_RAW.tar (≈113 MB, BEDPE + TXT)

The pipeline (CRSSANT), per repo README + paper Methods

  1. Preprocess FASTQ (adapter/barcode trim).
  2. STAR (v2.7.0f in paper) map reads — chimeric/softclip aware (--chimSegmentMin 5 --chimJunctionOverhangMin 5 --chimScoreJunctionNonGTAG 0 --chimOutType WithinBAM HardClip --outFilterMultimapNmax 10).
  3. softreverse.py — rearrange softclipped alignments, remap.
  4. gaptypes.py — classify primary alignments into 6 types: continuous / gap1 / gapm / trans / homo / bad.
  5. gapfilter.py — drop splice junctions + 1–2 nt gaps.
  6. crssant.py — cluster gap1+trans alignments into Duplex Groups (DGs) and Non-overlapping Groups (NGs). Params: t_o=0.5 (spectral)/0.1 (cliques), t_eig=5.
  7. gapmcluster.py — multi-gap → tri-segment groups (higher-order).

In scope (pipeline-derived, reproducible)

  • S1 Per-library % gapped reads — paper: "3.3–14.5% of the reads are gapped" (Results; per-library in Supplementary Table 1). Reproduce by mapping the SRA libraries and counting gap1+gapm vs total primary alignments.
  • S2 Duplex Groups (DGs) per RNA / per library — the deposited BEDPE files in GSE167812_RAW.tar ARE the crssant.py output. Reproduce DG assembly from the raw reads and compare DG counts to the deposited BEDPE (ground-truth pipeline output) and to any paper-stated counts (e.g. 7SK "17 DGs", Supp Fig 10f).
  • S3 Read/alignment classification counts (continuous/gap1/gapm/trans) per library — intermediate, checkable against Supplementary Table 1.

Reproduction strategy

Treat the deposited processed BEDPE/TXT (GSE167812_RAW.tar) as the authors' pipeline output. Two-tier:

  • Tier A (cheap, profiling): download the processed tar (small) → profile the BEDPE/TXT (N DGs, columns, RNAs covered) and cross-check against paper numbers.
  • Tier B (full repro, heavy → «our HPC»): download raw FASTQ for selected SRR (SRP308411) → STAR map → gaptypes → gapfilter → crssant.py → compare regenerated DG counts + %gapped against Tier-A deposited output and Supplementary Table 1.

Out of scope (not attempted)

  • Wet-lab: crosslinking-efficiency gels (1.01/1.31/1.89%), DD2D recovery, RNase titration — bench measurements, not pipeline-derived.
  • 3D-distance / structural-modeling claims (Å distances, 28S/18S 3D models, VARNA/ PyMOL renderings) — derived from external structures + manual modeling, not the read-processing pipeline. Noted but not reproduced.

Heavy-compute note

STAR human-genome mapping + alignment classification needs ≥30–100 GB scratch and RAM → must run on «our HPC» (SLURM). «host» only orchestrates. As of first pass the «our HPC» VPN tunnel is unreachable («host» ssh times out) — waiting for the central fix before submitting Tier-B jobs; Tier-A profiling can proceed once the tunnel is up (downloads happen on front1/«infra», never «host»).

Figures / tables: TableFig. 10f
T0a-gaptypes
Reported
10093 -> cont 9179, gap1 339, gapm 1, trans 1, homo 281, bad 0 (CRSSANT shipped test)
Reproduced
10093 -> cont 9179, gap1 339, gapm 1, trans 1, homo 281, bad 0
exact
T0b-gapfilter
Reported
352 -> 291 retained; 5 splice junctions (CRSSANT shipped test)
Reproduced
352 -> 291; filtered SAM byte-identical to shipped expected file
exact
T0c-crssant
Reported
24 DG rows (ACTB cliques t_o=0.1, shipped expected bedpe)
Reproduced
24 DGs byte-identical in 3/5 runs; 2/5 gave 25 DGs but still recovered 23/24 shipped DG coord-pairs. Clique clustering mildly non-deterministic (set/dict order).
within tolerance
T0d-gapmcluster-7SK
Reported
275 TG-tagged alignments (shipped RN7SK_hg38_gapm_tg.sam; human 7SK TG assembly, README Step 8)
Reproduced
275 alignments, output byte-identical to shipped expected file
exact
DS-N
Reported
12 deposited samples (GSE167812)
Reproduced
12 files present, parse cleanly
exact
S1-pct-gapped
Reported
3.3-14.5% of reads gapped (Supp Table 1)
Reproduced
deposited gapped-read counts per library characterized (13775-3239644); exact % needs Supp Table 1 denominators or per-library custom-genome remap (not run for all 10)
partial
S2-7SK-DGs
Reported
17 DGs on 7SK (Supp Fig 10f)
Reproduced
24-25 RN7SK DGs reproduced END-TO-END FROM RAW FASTQ (SRR13797237) via STAR->gaptypes->gapfilter->crssant on the 939 N-gapped both-arms-in-7SK reads (5 runs: 25/25/24/25/24). Top-17 DGs by read support (>=5 reads) form a well-supported core consistent with the reported 17; extra 7-8 are low-support (2-5 reads). Same order of magnitude; count is threshold/library/non-determinism dependent.
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 84/100

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)
🤝
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.

62.1 k
tokens (I/O) · 2.7 M incl. cache
20 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.