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

Widespread allele-specific topological domains in the human genome are not confined to imprinted gene clusters.

Genome Biol · 2023
L1 64/100 3/4
Why this verdict

The main results reproduced, with only marginal, non-material deviations.

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • Any deviation was negligible
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
  • 🟡A deviation was attributed to the published material
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
64/100
Reproducibility score
0.6 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 25% of all assessed papers rank 854 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 (honest). The authors' own HiCFlow Snakemake pipeline was reproduced end-to-end on a «our HPC» compute node, on its bundled Drosophila S2 example data (the demo the authors ship to validate the pipeline), via --use-conda (16 per-rule envs). It regenerated the complete documented output set (HiC matrices, OnTAD TADs, HiCcompare differential, viewpoint, HiCRep, insert-size, MultiQC) with concrete harvested numbers (HiCRep replicate SCC 0.963-0.978 — within the paper's RC1 range 0.97-0.98; OnTAD 11-17 TADs/sample; HiCExplorer Hi-C contacts 85-94%). Only ditagLength.svg + the optional fastQScreen contamination QC were skipped (no external genome DBs bundled; not a default target). NOT reproduced (documented blockers, not faked): the paper's headline human genome-scale numbers (99.8% genotyping / 99.9% phasing concordance, 39 conserved ASTADs, ASE/imprinted enrichment Z-scores) require regenerating EVERY HiCFlow intermediate from multi-TB GSE63525 across 3 cell lines + the AS-HiC-Analysis notebooks (which reference intermediates by external local paths and ship none) = node-weeks + multi-TB, out of bounded compute. RC1/RC2 on the paper's own PRJNA926951 RC-Hi-C is tractable in size but blocked by the missing capture-region BED.

💻 Code ↗ 🗄 Data: GSE63525

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.

✎ 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 2026-06-30
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 investigates whether allele-specific (parent-of-origin) differences in 3D chromatin conformation are limited to classical imprinted gene clusters, or whether allele-specific topologically associating domains (TADs) exist more widely across the human genome and relate to allele-specific gene expression.

Core claims
  • HiCFlow, a new bioinformatic pipeline, performs de novo haplotype assembly, phasing, and visualization of allele-specific (parental) chromatin conformation directly from Hi-C data without requiring pre-phased haplotypes. method
  • At the IGF2-H19 locus, known stable allele-specific CTCF-mediated looping interactions (ICR-dependent) are robustly detected in both RC-Hi-C and Hi-C datasets, consistent with loop-extrusion models. finding
  • SNRPN and DLK1 loci show more variable, less canonical allele-specific 3D structures than IGF2-H19, with allele-specific A/B compartmentalization detected instead of a single conserved imprinted structure. finding
  • Genome-wide unbiased ranking of TADs by allele-specific contact frequency identifies a defined set of allele-specific TADs (ASTADs), which occur in regions of high sequence variation. finding
  • ASTADs are enriched not only for imprinted loci but also for allele-specific expressed (ASE) genes genome-wide, including previously unreported loci such as the TAS2R bitter taste receptor gene cluster. finding
  • HiCFlow's Hi-C-based genotyping and haplotype phasing show high concordance (~99.8-99.9%) with an experimentally validated high-confidence haplotype in GM12878 cells. finding
  • Most allele-specific chromatin interactions occur within subTADs rather than defining entire TADs, and imprinted gene clusters share TADs with non-imprinted neighboring genes. finding
  • 8-32% of genes exhibiting allele-specific expression are located within ASTADs. finding
Experimental setups
Assay System Perturbation Readout Platform
Region Capture Hi-C (RC-Hi-C) 1-7HB2 human breast epithelial cell line none allele-specific chromatin contact frequency at imprinted loci (IGF2-KCNQ1, SNRPN, DLK1-DIO3) MboI 4-base cutter restriction enzyme, tiled capture probes
Hi-C (haplotype phasing benchmark) GM12878 human lymphoblastoid cell line none genotyping and phasing accuracy compared to high-confidence experimentally validated haplotype prototype haplotype-phased Hi-C data
Region Capture Hi-C / Hi-C IMR-90 and H1-hESC human cell lines none allele-specific interactions at imprinted loci
Genome-wide Hi-C / TAD analysis human cell lines (unspecified panel) none TAD insulation score, identification of allele-specific TADs (ASTADs) genome-wide
Allele-specific expression (ASE) analysis human cell lines none overlap of ASE genes with ASTAD locations
Key results
  • Known allele-specific enhancer interactions with IGF2 and H19 promoters robustly detected as opposing A1/A2 signals in the subtraction matrix at the IGF2-H19 locus.
  • HiCFlow genotyping agreed with the high-confidence GM12878 dataset for the vast majority of common variant calls. 99.8% agreement (n=3,799,226 common loci)
  • HiCFlow haplotype phasing agreed closely with the high-confidence dataset. 99.9% agreement (n=1,942,361 informative loci)
  • SNRPN locus shows weak/poorly defined TAD boundaries with directional allelic bias in long-range associations (A2 biased leftward, A1 biased rightward).
  • DLK1 locus subTAD structure differs by allele: A1 forms a larger subTAD anchored upstream of DLK1, while A2 subTAD is anchored at the imprinting control region (ICR).
  • A defined fraction of genes with allele-specific expression are located within genome-wide-identified ASTADs. 8-32%
  • RC-Hi-C library generated high sequencing depth and coverage comparable to published high-resolution Hi-C datasets. ~40 million valid read pairs, ~1700 read pairs/kb
Key statistics
  • other 99.8% agreement in variant identity (GM12878 HiCFlow genotyping vs. high-confidence dataset)
  • other 99.9% agreement in phasing (GM12878 HiCFlow haplotype phasing vs. high-confidence dataset)
  • count 4,267,624 (HiCFlow) vs 4,049,512 (high-confidence) variants identified (GM12878 genotyping comparison)
  • count 2,147,688 (HiCFlow) vs 2,063,320 (high-confidence) phased variants identified (GM12878 phasing comparison)
  • count 34,399 probes covering ~4.1Mb (~16.1%) of capture regions (RC-Hi-C probe design across 5 imprinted loci (25Mb total))
  • count ~40 million valid read pairs, mean coverage ~1700 read pairs per kilobase (RC-Hi-C sequencing depth in 1-7HB2 cells)
  • other 8-32% (proportion of allele-specific expressed genes located within ASTADs)

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 paper describes a bioinformatic pipeline (HiCFlow) for haplotype phasing and allele-specific analysis of Hi-C/Region Capture Hi-C data, applied to imprinted gene loci and genome-wide TAD calling. Results are reported largely as descriptive comparisons (visual subtraction matrices, percent concordance between genotyping/phasing pipelines, and qualitative/enrichment observations about allele-specific TAD distributions) rather than through a dedicated inferential statistics section. Note: the supplied text is a partial excerpt (cuts off before a Methods/Statistics section), so this assessment is based only on what is shown.

Replicationunclear GroupsAllele 1 (A1) vs Allele 2 (A2) chromatin interaction/contact frequencies at imprinted loci and genome-wide TADs, across multiple cell lines (1-7HB2, IMR-90, H1-hESCs, GM12878) Pairingunclear Randomization/blindingnot stated Dispersionnone
Approaches that could also have been used
  • Allele-specific TADs (ASTADs) were defined by unbiasedly ranking TADs according to their allele-specific contact-frequency differences.
    Could also: A formal statistical/count-based test for differential chromatin interactions (e.g., paired permutation testing, or Hi-C-adapted count models such as diffHic or multiHiCcompare) could also be applied to each TAD. — This would pair the ranking with a significance threshold and false-discovery-rate control, letting readers distinguish TADs with strong statistical support from those near the top of the ranking by chance.
  • Genotyping and phasing concordance between HiCFlow and the high-confidence reference dataset was reported as single percentage values (99.8% and 99.9% agreement).
    Could also: Reporting these proportions alongside a confidence interval (e.g., Wilson or Clopper-Pearson) or a chance-corrected agreement statistic such as Cohen's kappa would also be a standard way to summarize concordance. — A CI or kappa statistic conveys the precision of the concordance estimate given the number of loci compared, complementing the raw percentage.
  • Enrichment of allele-specific TADs in regions of high sequence variation and for allele-specific expressed genes was described narratively.
    Could also: A formal enrichment test (e.g., Fisher's exact test, hypergeometric test, or permutation-based enrichment tools like regioneR/GREAT) with multiple-testing correction could also be used. — This would provide a quantitative estimate (odds ratio or fold enrichment) and an FDR-adjusted p-value indicating how unlikely the observed overlap is under a null model of random genomic overlap.
  • Differences between alleles at individual loci (e.g., IGF2-H19, SNRPN, DLK1) were assessed via visual inspection of subtraction matrices denoised with a median filter.
    Could also: A quantitative, bin-pair-level statistical comparison of matched contact counts between alleles (e.g., using Hi-C-specific differential interaction tools such as diffHic, HiCcompare, or a paired non-parametric test on normalized counts) could also complement the visualization. — This would let each highlighted allelic difference be accompanied by an effect size and significance value rather than relying solely on visual/color-based interpretation of the subtraction matrix.
  • Variation in ASTAD distribution across cell lines was described qualitatively ("The ASTAD distribution varied between cell lines").
    Could also: A formal statistical comparison of TAD-count proportions across cell lines (e.g., a chi-square or Fisher's exact test on contingency counts) could also be used. — This would give a quantitative basis (test statistic and p-value) for the claim of between-cell-line variability, in addition to the descriptive comparison.
Software: HiCFlow (custom pipeline developed in this study) · HiC-Pro (mentioned as an existing alternative pipeline, not used for primary analysis)

What was reproduced

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

Scope — pmid-36869353

Paper: Richer S, Tian Y, Schoenfelder S, Hurst L, Murrell A, Pisignano G. "Widespread allele-specific topological domains in the human genome are not confined to imprinted gene clusters." Genome Biol 2023. PMCID PMC9983196, DOI 10.1186/s13059-023-02876-2.

Code (MIT):

Data:

  • Public Hi-C: GM12878, IMR-90, H1-hESC in-situ Hi-C — GEO GSE63525 (Rao et al. 2014; 200 samples, 4.9 billion contacts in GM12878) / also distributed via 4D Nucleome.
  • Generated RC-Hi-C (1-7HB2 / HB2 cells): NCBI PRJNA926951 (2 runs SRR23199640 HB2_WT2, SRR23199639 HB2_WT1; Hi-C; ~175M + ~161M read pairs).
  • Auxiliary (not pipeline-derived here): CTCF ChIP (ERX115548 + ENCODE), Roadmap chromHMM 15-state, ASMdb allele-specific methylation, geneimprint.com imprinted genes, Peakachu loops (3D Genome Browser), GTEx eQTL, GWAS catalog.

Pipeline-derived results (candidate reproduction targets)

id result reported value paper loc pipeline feasibility
EX1 HiCFlow bundled example (Wang 2018 Drosophila S2 cells) reproduces the documented HiC/HiCcompare/viewpoint/HiCRep/MultiQC outputs "produces the exact figures as shown" (README) HiCFlow README HiCFlow (whole) HIGH — authors' own code + bundled ~550 MB data; bounded compute
KT1 Karyotype / ploidy & CNV per chromosome (GM12878/IMR90/H1hESC) aneuploidy/CNV calls (Add. file, Fig CNVstatus) AS-HiC 0.checkKaryotype, 0.processCNV nQuire + binned matrices (.bin shipped) MED — only step whose inputs are bundled in the repo
PH1 Genotyping concordance GM12878 Hi-C vs GIAB high-confidence 4,267,624 variants called vs 4,049,512 truth; 3,799,226 common; 99.8% agreement Results / Fig (phasing) HiCFlow CallVariant (GATK) LOW — needs full GM12878 raw Hi-C (multi-TB)
PH2 Phasing concordance GM12878 vs GIAB 2,147,688 phased vs 2,063,320; 99.9% over 1,942,361 loci Results HiCFlow Phase (HapCUT2/SNPsplit) LOW — same blocker
AT1 # conserved ASTADs across 3 cell lines (90% reciprocal overlap) 39 conserved ASTADs; 5 imprinted genes within Results / Fig AS-HiC 2.summariseASTAD,8.overlapAnalysis LOW — needs all 3 cell lines fully processed; repo ships no intermediates
AT2 ASE enrichment in ASTADs GM12878 153/480 (32%) Z=4.64 p=1.7e-6; IMR90 73/409 (18%) Z=3.43 p=3e-4; H1 182/2398 (8%) Results AS-HiC 1.ASE_and_Imprinted randomisation LOW — same blocker
AT3 Imprinted-gene enrichment in ASTADs H1 45/115 (39%); IMR90 38 (33%); GM12878 42 (37%); p≤0.001 Results AS-HiC 1.ASE_and_Imprinted LOW — same blocker
RC1 RC-Hi-C reproducibility between WT replicates 0.97–0.98 at 5 kb (HiCRep) Results HiCFlow on PRJNA926951 MED — paper's own ~20 GB data; needs capture-region BED

In scope (attempted)

  • EX1 (primary, clean 1:1 of the authors' pipeline on bundled data) — run on «our HPC».
  • KT1 (stretch) — repo-bundled .bin matrices → nQuire ploidy.
  • RC1 (stretch) — PRJNA926951 RC-Hi-C HiCRep, if capture BED resolvable.

Out of scope / not attempted (honest blockers)

  • PH1/PH2/AT1/AT2/AT3 — the headline genome-scale numbers. Reproducing them requires regenerating every HiCFlow intermediate (aligned BAMs, GATK VCFs, HapCUT2 phases, SNPsplit allele matrices, OnTAD ASTAD calls) from the multi-TB GSE63525 raw Hi-C for all three cell lines, then running the AS-HiC-Analysis notebooks. The AS-HiC-Analysis repo references those intermediates by local/external paths (../../GM12878/..., /media/stephen/Elements/...) and ships none of them except the karyotype .bin matrices. Thi
EX1
Reported
HiCFlow bundled Drosophila S2 example produces the documented HiC/HiCcompare/viewpoint/HiCRep/MultiQC outputs (README)
Reproduced
REPRODUCED: authors' Snakemake pipeline ran end-to-end on «our HPC» via --use-conda (16 per-rule envs, RUNRC=0); regenerated the complete documented output set (HiC matrices+OnTAD TADs, HiCcompare diff png, viewpoint svgs, HiCRep svg, insertSize svg, pyGenomeTracks, FastQC+HiCExplorer QC, MultiQC). Only ditagLength.svg absent (not a default-config target).
within tolerance
EX1-hicrep
Reported
HiCRep replicate reproducibility (demo; ties to RC1 0.97-0.98)
Reproduced
SCC G1S-1vsG1S-2=0.978, AS-1vsAS-2=0.963; cross-condition 0.886-0.896
within tolerance
PH1
Reported
GM12878 Hi-C: 4,267,624 variants vs 4,049,512 GIAB; 99.8% agreement
Reproduced
NOT_ATTEMPTED (multi-TB raw + full GATK; out of bounded compute)
partial
AT1
Reported
39 conserved ASTADs across 3 cell lines
Reproduced
NOT_ATTEMPTED (needs all regenerated intermediates; repo ships none)
partial
RC1
Reported
RC-Hi-C reproducibility 0.97-0.98 at 5kb (HiCRep)
Reproduced
NOT_FULLY_ATTEMPTED (capture BED not locatable); method demonstrated via EX1-hicrep (0.96-0.98)
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 64/100

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.

79.8 k
tokens (I/O) · 5.5 M incl. cache
11 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.