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

snpQT: flexible, reproducible, and comprehensive quality control and imputation of genomic data.

F1000Res · 2021
L1 67/100 3/4
⚑ Flagged for review — a reproduced result did not match the reported value

Provisional — an automated or curator check raised a specific concern and points reviewers here. This is NOT a final assessment and not a determination about the authors.

Why this verdict

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

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • Same input data as the authors
  • Reported values were directly comparable
What did not (or only partly)
  • 🟡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 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
67/100
Reproducibility score
0.4 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 29% of all assessed papers rank 795 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 (3 of 4 pipeline tiers reproduced, 2 of them with 1:1 figure matches). snpQT is a well-documented Nextflow GWAS QC+imputation pipeline (authors' own code wrapping third-party tools); reproduced on the shipped synthetic toy dataset (only open input; ALS is dbGaP-restricted). On «our HPC» via SLURM + conda profile («job»): (1) --qc reproduced 1:1 -- variant QC 6515->2612, sample QC 100->48, variant_qc_log figure visually IDENTICAL to the repo reference; (2) --pop_strat reproduced 1:1 -- PCA against 1000G with toy samples overlaying EUR, PC1vsPC2_before.png visually IDENTICAL to reference (only legend label pop vs race differs); (3) --gwas association reproduced -- plink2 logistic --glm with/without 3 PC covariates on 48 samples/2612 variants/22-26 case-control, lambda 1.058/1.0, top hit rs6682769, though the R Manhattan/QQ figure step failed on forcats version-rot. NOT reproduced: imputation (headline 6,500,533) -- conda blocks it and impute5 is license-gated (the container runtime itself now works via apptainer, so this corrects the prior 'no containers' finding). Toy composition shows paper-text-vs-shipped-data drift (6515 vs 6517 variants; 48/52 vs 51/49 case/control at commit b4b570a) -- the authors' OWN reference figure also starts at 6515, so this is post-publication data regeneration, not a pipeline error. Required engineering fixes were all dependency-rot (NXF_SYNTAX_PARSER=v1, conda.enabled, «infra»-prefix envs, snpflip setuptools/natsort rebuild), not science changes. Did NOT attempt: imputation (license blocker) and ALS benchmarks (restricted data).

💻 Code ↗ 🗄 Data: 10.5281/zenodo.5703398

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 65
    assessed: 2026-06-20 ⛓ 677859a26418
✎ 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
2026-06-22
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

Existing semi-automated or automated genomic quality control pipelines lack comprehensive quality checks, flexible workflow architecture, and user control, so a scalable, containerized nextflow-based pipeline (snpQT) can provide more comprehensive, reproducible, and flexible QC and imputation of human genomic data.

Core claims
  • snpQT is a scalable, stand-alone software pipeline using nextflow and BioContainers for comprehensive, reproducible, interactive QC of human genomic data. resource
  • snpQT offers some 36 discrete quality filters or correction steps in a complete standardised pipeline. method
  • snpQT includes human genome build conversion, population stratification against 1,000 Genomes Project data, automated population outlier removal, and built-in imputation with its own pre- and post-QC. method
  • snpQT is organized as nine core workflow components (A-I) implemented with nextflow, executable via Docker, Singularity, Anaconda, or Environment Modules. method
  • Nextflow caching allows skipping of unchanged earlier pipeline stages when parameters are altered later in the workflow, saving time. mechanism
  • Each main workflow (C, D, E, I) provides a 'Make Report' module generating interactive HTML reports of before/after threshold plots. method
  • snpQT is open source (MIT license) and freely available at https://github.com/nebfield/snpQT, with a synthetic dataset and online tutorial provided. resource
  • snpQT supports both local imputation (shapeit4/impute5) and export for use with external imputation servers or alternative reference panels. method
Experimental setups
Assay System Perturbation Readout Platform
SNP-array genotype QC and pipeline demonstration (snpQT synthetic dataset use case) synthetic human genomic dataset, chromosome 1, 100 female samples (51 cases, 49 controls) none (demonstration run of snpQT QC pipeline) 6,517 genotypes; QC plots and reports before/after thresholds snpQT (nextflow) with PLINK, PLINK2, picard, samtools, snpflip
GWAS pipeline QC, imputation, and association testing (real-world tutorial dataset) Amyotrophic Lateral Sclerosis SNP-array dataset, 2,000 samples (1,000 cases, 1,000 controls) from a restricted-access dbGaP project none (pipeline QC/imputation/GWAS applied) quality-filtered samples/variants, Manhattan plot, Q-Q plot, GWAS summary statistics snpQT with shapeit4 and impute5
Population stratification / outlier removal via PCA user dataset merged with 1,000 Genomes Project reference panel automated outlier removal (EIGENSOFT) principal components before and after outlier removal EIGENSOFT
Human genome build conversion user genomic dataset (VCF/PLINK format) liftover between GRCh38 and GRCh37 converted genomic coordinates Picard LiftoverVcf
Key results
  • snpQT implements approximately 36 discrete QC filters/correction steps across nine workflows. 36 steps
  • Standard QC checks require less than 16GB RAM and 4 cores on large datasets of up to 40,000 individuals. <16GB RAM, 4 cores / 40,000 individuals
  • Local imputation requires substantially more computing power than standard QC. up to 50GB RAM per chromosome per core
  • Core reference database download and storage size for snpQT setup. 17.3GB compressed / 19.7GB unzipped
  • Imputation reference database download and storage size. 13GB compressed / 15GB unzipped
  • Synthetic demonstration dataset provided with defined structure for reproducible testing. 6,517 genotypes, 100 samples, 51 cases/49 controls
Key statistics
  • count 36 QC filters/correction steps (total quality filters/correction steps implemented in snpQT)
  • count 9 core workflows (A-I) (number of nextflow workflow components in snpQT architecture)
  • count 6,517 randomised genotypes, 100 female samples (51 cases, 49 controls) (synthetic demonstration dataset composition)
  • count 2,000 samples (1,000 cases, 1,000 controls) (real-world ALS SNP-array dataset used in online tutorial)
  • other 17.3GB compressed / 19.7GB unzipped (core reference database size)
  • other 13GB compressed / 15GB unzipped (imputation reference database size)
  • other <16GB RAM and 4 cores (hardware requirement for QC on datasets up to 40,000 individuals)
  • other up to 50GB RAM per chromosome per core (hardware requirement for imputation)

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 tool article describing snpQT, a Nextflow-based pipeline for GWAS quality control and imputation; it does not report a primary study with inferential statistical comparisons between experimental groups. Statistical content is descriptive of the QC and GWAS methodology the pipeline implements (HWE tests, PCA-based population stratification, imputation, association analysis), demonstrated on a synthetic 100-sample dataset and a referenced ALS dataset. No hypothesis tests are applied by the authors to evaluate or benchmark the tool itself; performance is assessed narratively through workflow outputs and added performance metrics for two cohort sizes.

Replicationunclear Sample sizeSynthetic demonstration dataset: 100 female samples (51 cases, 49 controls), 6,517 randomised chromosome 1 genotypes; real-world ALS dataset referenced as 2,000 samples (1,000 cases / 1,000 controls) from dbGaP — used only for online tutorial, not formally analysed in the paper GroupsCases vs. controls within the synthetic/ALS datasets (as inputs to the pipeline); no statistical comparison performed by the authors between groups Pairingna Randomization/blindingnot stated Dispersionnone Multiplicity correctionnot stated
Statistical tests used
Test Applied to n Assumptions
Hardy-Weinberg Equilibrium exact test (implemented via PLINK as a pipeline step) Workflow E: Variant QC — filtering SNPs deviating from HWE not stated
Principal Component Analysis (PCA) for population stratification and outlier removal (via EIGENSOFT) Workflow D: Population Stratification — automated outlier removal against 1000 Genomes reference not stated
Logistic/linear regression-based GWAS association analysis (implemented via PLINK/PLINK2.0 with PC covariates) Workflow I: GWAS with and without covariate adjustment; outputs Manhattan plot and Q-Q plot not stated
Identity-by-descent (IBD) estimation for cryptic relatedness (via PLINK) Workflow C: Sample QC — removal of cryptically related individuals not stated
Haplotype phasing (SHAPEIT4) and genotype imputation (impute5) Workflows F–H: pre-imputation QC, local imputation, post-imputation QC not stated
Approaches that could also have been used
  • Population stratification is handled with PCA via EIGENSOFT, with automated outlier removal relative to 1000 Genomes Project reference clusters
    Could also: Linear mixed models (e.g. BOLT-LMM, SAIGE, REGENIE) or ADMIXTURE-based ancestry estimation could also be used to control for population structure — Mixed-model approaches jointly model population structure and relatedness in the association step itself, which can offer improved control in admixed or family-based cohorts and may be preferred when population stratification and cryptic relatedness co-occur at scale
  • Cryptic relatedness is assessed via IBD estimation in PLINK, with related individuals removed
    Could also: KING (Kinship-based INference for GWAS) is also widely used for relatedness estimation, particularly in large or diverse datasets — KING uses a moment estimator that is more robust to population structure than PLINK's method-of-moments IBD, which can be advantageous in ethnically diverse cohorts
  • Local imputation is performed using SHAPEIT4 for phasing and impute5 for imputation against a 1000 Genomes reference panel
    Could also: Beagle 5.x or Michigan Imputation Server (using HRC or TOPMed panels) are also commonly used; the pipeline already notes compatibility with external imputation servers — TOPMed-based imputation offers a much larger and more diverse reference panel, potentially yielding better imputation quality for rare variants and non-European ancestry populations
  • GWAS association analysis is implemented with logistic/linear regression adjusted for principal components as covariates (via PLINK/PLINK2.0)
    Could also: Mixed-model association methods (BOLT-LMM, SAIGE, REGENIE) could also be applied, incorporating a genetic relationship matrix to control for both population stratification and sample relatedness simultaneously — Mixed-model methods are particularly well-suited to biobank-scale datasets and cohorts with population structure or relatedness, where simple PC-covariate adjustment may be insufficient
  • Post-imputation QC removes poorly imputed variants but the specific imputation quality metric threshold (e.g. INFO score) is user-modifiable without a single recommended default stated in the manuscript
    Could also: Reporting and filtering on both INFO score (impute5) and Rsq (minimac) with thresholds stratified by minor allele frequency is also a standard approach, as recommended by the TOPMed and HRC imputation guidelines — Frequency-stratified imputation quality thresholds acknowledge that rare variants are harder to impute accurately, allowing differential stringency and potentially retaining more well-imputed common variants while excluding poorly imputed rare ones
  • The synthetic validation dataset is small (100 samples, 6,517 SNPs on chromosome 1) and artificially structured; a real-world dataset is demonstrated only in the online tutorial
    Could also: A formal benchmarking comparison against existing pipelines (e.g. GenotypeHarmonizer, ricopili, or the Anderson et al. tutorial scripts) on a shared public dataset could also be used to evaluate tool performance — Head-to-head benchmarking on common datasets would enable users to directly compare QC outputs, runtime, and reproducibility across tools, supporting evidence-based tool selection
Software: Nextflow v21.04.3 · PLINK · PLINK2.0 · EIGENSOFT · SHAPEIT4 · impute5 · Picard (LiftoverVcf) · samtools · R · Anaconda / BioContainers / Singularity / Docker

What was reproduced

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

Figures / tables: figures
C1_toy_n_samples
Reported
100 samples
Reproduced
100
exact
C2_toy_sex
Reported
all female (100)
Reproduced
100 sex=2 (female); gwas log 48 females/0 males
exact
C3_toy_n_variants
Reported
6,517 chr1 genotypes
Reproduced
6,515
did not match
C4_toy_case_control
Reported
51 cases / 49 controls
Reproduced
48 cases / 52 controls (raw); 22/26 post-QC
did not match
C5_toy_chromosome
Reported
chromosome 1
Reproduced
all chr1
exact
C6_qc_pipeline
Reported
per-step sample/variant QC log figures
Reproduced
Variant QC 6515->2612(HWE)->2612; Sample QC 100->48; variant_qc_log figure visually IDENTICAL to repo reference
exact
C9_pop_strat
Reported
PCA vs 1000G superpopulations (before/after figures)
Reproduced
pop_strat end-to-end; merged w/ 2504 1000G (228 variants x 2552 samples); 0 toy outliers removed; PC1vsPC2_before.png visually IDENTICAL to repo reference (legend pop vs race)
exact
C10_gwas
Reported
Manhattan + Q-Q plots of logistic GLM (with/without covariates)
Reproduced
association reproduced: plink2 --glm firth-fallback, 48 samples/2612 variants/22 cases/26 controls; lambda=1.05764 (3 PC covars), 1.0 (no covars); 2612 tests; top hit rs6682769 UNADJ=0.00138. Manhattan/QQ/log FIGURES failed (gwas:parse_logs forcats version-rot)
partial
C7_imputation_variants
Reported
6,500,533 variants on chr1 after imputation
Reproduced
NOT REPRODUCED (license-gated impute5)
partial
C8_als_runtime
Reported
ALS 2k: 4min; 12,319: 3h14m
Reproduced
OUT OF SCOPE (dbGaP-restricted)
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 67/100

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.

713.7 k
tokens (I/O) · 77.3 M incl. cache
293 min
runtime · 0.14 CPU-h
4.8 GB
peak RAM
4
HPC jobs
hummel
machine