Enriched domain detector: a program for detection of wide genomic enrichment domains robust against local variations.
The main results reproduced: recomputed values matched the published ones within tolerance.
Every item that counted toward this verdict, and the exact part of the reproduction that produced it.
- ✓Same input data as the authors
- ✓Reported values were directly comparable
- ✓No authors-side cause for any deviation
- ✓Reported values are derivable from the shared data
- ✓Any deviation was negligible
- ✓The central claim held under reproduction
- ✓Overall, the reproduction was clean
- 🟡A deviation arose in the data or preprocessing
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
2/3 replicates reproduced exactly on paper spec (within ~2%); rep1 required EDD's own documented auto-bin fallback (4Kb vs paper's 3Kb) due to EDD's internal informative-bins QC, graded partial for peak count, within-tol for coverage. Growing-condition LMNB1 ChIP fully reproduced; senescent condition + IgG controls out of scope for this pass.
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.
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
- 2026-08-01
- Rubric version
- v1.0
- Assessed by
-
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-08-01no human curator yet
- Last updated
- 2026-08-01
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: opusExisting broad ChIP-seq peak callers are not suited to detect the megabase-size, low-level, locally variable enrichment domains through which nuclear lamins contact the genome; the paper tests whether a new algorithm (enriched domain detector, EDD) based on ChIP/input bin ratios, logit bin scoring with a gap penalty, maximal scoring subsequences and Monte Carlo significance testing can identify such lamina-associated domains robustly.
- ★ EDD is a new algorithm that detects broad (megabase-size) enrichment domains from ChIP-seq data of widely distributed chromatin proteins such as A- and B-type lamins. method
- ★ EDD's advantage over existing broad peak callers is sensitivity to domain width rather than enrichment strength at a particular site, plus robustness against local variations. finding
- ★ EDD models the ChIP-to-input relationship per bin (p-hat = ChIP reads / (ChIP + input reads)) and selects the smallest bin size yielding at least 99% 'informative' bins, classifying bins as enriched (EIB), depleted (DIB) or non-informative (NIB). mechanism
- ★ Bins are scored with a logit function, with depleted/non-informative bins downweighted by a user- or auto-selected gap penalty (auto-optimizing EPR^5 * ECR), and peaks are found as maximal scoring subsequences with Monte Carlo-based significance testing. mechanism
- ★ EDD can discover genomic domains enriched in lamin A using new LMNA ChIP-seq data. finding
- ★ EDD is released as an open Python package (PyPI) with source code and manual at http://github.com/CollasLab/edd, and new LMNA ChIP-seq data are deposited under GEO GSE54334. resource
- RSEG is too lenient in a lamin context: lamin domains it identifies have very large genome coverage with many domains of very low enrichment. finding
- Existing callers are limited for lamins because BroadPeak lacks input support, while SICER and PeakRanger call putative peaks from ChIP data alone and only use input afterwards to assess significance. finding
| Assay | System | Perturbation | Readout | Platform |
|---|---|---|---|---|
| ChIP-seq (lamin A/C ChIP followed by high-throughput sequencing) | Human normal dermal fibroblasts (Lonza CC-2511, LDF) and human normal primary dermal fibroblasts (AD04DF), passage 5-7, harvested at confluency | none (anti-lamin A/C antibody Santa Cruz sc-7292 vs irrelevant mouse IgG control) | Genome-wide LMNA-associated sequence read counts / enriched domains (LADs) | Illumina HiSeq2500; Bioruptor (Diagenode) sonication; Dynabeads Protein G (Invitrogen); Illumina library prep |
| ChIP-qPCR (lamin A) | Human dermal fibroblasts (LMNA ChIP DNA) | none | Relative enrichment of specific loci in LMNA ChIP DNA (primers in Supplementary Table S2) | BioRad MyiQ Real-time machine with SYBR Green |
| RNA-seq | Human dermal fibroblasts | none | Transcript levels / gene expression (29.3 × 10^6 reads aligned) | Ambion TRIzol RNA extraction kit (Life Technologies); Illumina library; Illumina HiSeq2500; TopHat/Cufflinks |
| Computational ChIP-seq peak calling / benchmarking of EDD v1.0 against PeakRanger v1.17 (CCAT), SICER v1.1, BroadPeak and RSEG v0.4.8 | In silico analysis of LMNA (this study) and published LMNB1 ChIP-seq data sets with matched input | none (parameter sweeps; scripted testing of a range of parameter values) | Detected enrichment domains, genome coverage and domain enrichment levels | Bowtie2 v2.1.0 (HG19), Picard MarkDuplicates and DownsampleSam v1.86, IGV, Illumina iGenomes UCSC HG19 annotation |
| Re-analysis of published LMNB1 ChIP-seq and matched input sequence data | Public data set (NCBI GEO GSE49341) | none | LMNB1-enriched genomic domains | EDD (3 Kb bins, gap penalty 12) and comparator peak callers |
- – EDD enables quantitative analysis of ChIP-seq data for proteins that are widely distributed and show low-level enrichment on chromatin, outperforming existing broad peak callers for lamin domains.
- – EDD discovers genomic domains enriched in LMNA from new LMNA ChIP-seq data in human dermal fibroblasts.
- ▲ RSEG-identified lamin domains show very large genome coverage with many domains of very low enrichment, i.e. RSEG is too lenient for lamin data.
- – Extreme p-hat values (close to 0 or 1) occur almost exclusively in bins with few reads and wide confidence intervals; increasing bin size so that p-hat estimates p well in most bins makes downstream analysis more robust against noise.
- – For BroadPeak, subtracting input read counts from ChIP read counts per bin gave the most convincing results compared with using ChIP reads alone.
- – Keeping duplicate reads does not significantly affect LAD detection, owing to the large size of LADs (Supplementary Table S1).
- – The auto-selection function EPR^5 * ECR best predicted the manually selected gap penalties across the data sets analyzed during EDD development and testing.
- other 95% confidence interval for p-hat computed by the Agresti-Coull method; confidence interval threshold 0.25 defines 'informative bins' (EDD bin classification criterion)
- other at least 99% informative bins (excluding bins without reads) (Criterion for EDD selecting the smallest bin size)
- other p-hat > 0.5 (EIB, enriched) vs p-hat <= 0.5 (DIB, depleted) (Bin enrichment classification cutoff)
- other bin size 3 Kb with gap penalty 12 (LMNB1); bin size 11 Kb with gap penalty 5 (LDF, AD04DF_rep1) and 4 (AD04DF_rep2) for LMNA (EDD v1.0 parameters used)
- count 29.3 × 10^6 reads (RNA-seq reads aligned with TopHat/Cufflinks)
- other effective genome size 0.77 (LMNA, 51 bp reads) and 0.72 (LMNB1, 36 bp reads); window 200 bp, gap 600 bp, fragment size 300 bp, FDR cutoff 0.1 (SICER v1.1 parameters)
- other window size 500 bp, step size 50 bp (PeakRanger v1.17 CCAT parameters for broad peak detection)
- count 10^7 cells per ChIP; 50 μg lamin A/C antibody; chromatin fragments of ~200-400 bp; LADs typically 0.1-10 Mb (LMNA ChIP-seq experimental scale and LAD size range)
Statistical methods review
Model: sonnetA neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.
The paper describes EDD, a new algorithm for detecting broad genomic enrichment domains from ChIP-seq data, and benchmarks it against existing broad-peak callers (SICER, PeakRanger, BroadPeak, RSEG) using published and newly generated LMNA/LMNB1 ChIP-seq data. EDD's core statistical components include a 95% confidence interval for the binomial ChIP/input read-ratio (Agresti–Coull method) used to classify genomic bins as informative, a logit-based bin scoring scheme, and a Monte Carlo simulation (permutation of bin order across the genome) used to assess significance of candidate enriched domains. Comparator peak callers were run with default or manually tested parameters and results were assessed by visual inspection in a genome browser. ChIP-qPCR validation was performed in duplicate reactions, and RNA-seq reads were aligned with Cufflinks/TopHat using default parameters.
| Test | Applied to | n | Assumptions |
|---|---|---|---|
| 95% confidence interval for a binomial proportion (Agresti–Coull method) | classification of genomic bins as 'informative' vs 'non-informative' based on the ChIP/input read-count ratio p-hat | number of ChIP + input reads per bin (bin size chosen so that ≥99% of bins are informative) | not stated |
| Monte Carlo simulation (permutation of bin order) to derive an empirical significance measure | significance testing of candidate enriched domains (maximal scoring subsequences) identified by EDD | — | not stated |
| FDR cutoff (0.1) | peak calling by the comparator algorithm SICER during benchmarking against EDD | — | na |
-
Confidence intervals for the bin-level ChIP/input ratio (p-hat) were computed using the Agresti–Coull method for binomial proportions.↳ Could also: Wilson score interval or Clopper-Pearson 'exact' interval — these are also standard methods for constructing confidence intervals for binomial proportions and can behave somewhat differently at low counts or extreme proportions, offering an alternative way to define 'informative' bins.
-
Significance of candidate enriched domains was assessed via a Monte Carlo simulation that shuffles bin order across the genome and compares the observed top-scoring subsequence to this null distribution.↳ Could also: A parametric count-based model, such as negative-binomial modeling of ChIP versus input read counts (as used in tools like DESeq2 or MACS2) — a parametric approach could also be used and may offer faster computation and model-based confidence estimates, potentially complementing a permutation-based null distribution.
-
Bin size was selected as the smallest size producing at least 99% 'informative' bins, a threshold tuned per dataset.↳ Could also: A fixed, pre-specified bin size across datasets, or a multiscale/sliding-window analysis evaluating several bin sizes together — either could also be used and may support comparability of results across independent datasets by removing a dataset-specific tuning step.
-
Comparator peak callers (SICER, PeakRanger, BroadPeak, RSEG) were run across parameter ranges with suboptimal outputs excluded by visual inspection in a genome browser.↳ Could also: A quantitative benchmarking metric, such as precision/recall against a reference domain set or a reproducibility statistic (e.g., irreproducible discovery rate) across replicates — quantitative metrics could also provide an objective, reproducible basis for comparing algorithms alongside visual inspection.
-
SICER benchmarking used a single fixed FDR threshold (0.1) for calling peaks.↳ Could also: Reporting outcomes across a range of FDR thresholds, or applying a shared multiple-testing correction (e.g., Benjamini–Hochberg) consistently across all benchmarked callers — this could also let readers see how sensitive the benchmarking comparison is to the chosen significance threshold.
-
ChIP-qPCR validation was performed using duplicate reactions from the same chromatin preparation.↳ Could also: Additional independent biological replicates (separate chromatin preparations) alongside technical qPCR duplicates — biological replication could also be used to capture variability arising from chromatin preparation and ChIP efficiency, complementing the precision captured by technical replicate duplicates.
What was reproduced
The exact results taken into scope, with each reported value next to the value our attempt produced.
Assessments & scoring basis
Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.
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.
Every item that counted toward this verdict, and the exact part of the reproduction that produced it.
Input identity is clean: all 18 GSM samples of the public GSE49341 were obtained and the authors' own EDD tool (CollasLab/edd) was run, so reported and reproduced endpoints are directly comparable. Under the paper's exact spec (--bin-size 3 -g 12) rep2 gave 1928 vs ~1890 domains and rep3 1847 vs ~1815, with coverage within +0.3%/+0.5%; rep1 deviated more (1648 vs ~1803 domains, -8.6%) only because EDD's own informative-bins QC rejected the 3Kb bin and fell back to its documented 4Kb auto-estimate, yet even there coverage matched to +2.3% (1065.4 vs ~1041.675 Mb). The deviation therefore sits on the input/binning side and is caused by expected tool-internal behaviour — not by an authors' defect and not by a self-chosen methodological shortcut. Scope was narrowed (senescent d6-4OHT condition and 6 IgG controls not processed), but that limits breadth of coverage, not the validity of the claims that were tested; the central quantitative conclusion holds fully.
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.