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

A HML6 endogenous retrovirus on chromosome 3 is upregulated in amyotrophic lateral sclerosis motor cortex.

Sci Rep · 2021
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
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

REPRODUCED (1:1 on the public replication claim, via the paper's own named tool ERVmap). The paper's central finding — the HML6_3p21.31c provirus (ERVmap locus id 943, chr3:46426676-46433564 hg38) is upregulated in ALS motor cortex — reproduces on independent open data (GSE124439, the public motor-cortex subseries of GSE137810). Pipeline = ERVmap (bwa mem hg38 -> parse_bam.pl filter -> bedtools coverage -sorted over ERVmap.bed) for ERV counts + STAR GeneCounts for DESeq2 size factors + DESeq2 ~condition. For id943 (37 ALS/4 ctrl lateral motor cortex): log2FC=+3.456, p=0.00323, ranking #9 by p-value among ERVmap loci. DIRECTION matches the paper exactly and the replication p-value matches almost exactly (0.00323 vs reported 0.003). The log2FC MAGNITUDE is ~5x the reported 0.642 — expected given the different cohort composition (37/4 public vs 32/6 reported), absence of the paper's covariates in public GEO, and only 4 controls. This was a fully fresh re-run from raw FASTQ; the DESeq2 output is byte-identical to a prior independent run (determinism confirmed). NOT ATTEMPTED: the restricted KCL discovery numbers (data on-request), the full covariate model, and genome-wide multiple-testing / TEtranscripts / secondary GWAS-eQTL analyses. All 41 public samples processed; all in-scope claims graded. The BRIEF 'Code' link (R-Finance/PerformanceAnalytics) is spurious/unrelated — the paper's actual tool is ERVmap (github.com/mtokuyama/ERVmap), a valid P16 third-party-tool reproduction.

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 84
    assessed: 2026-06-20 ⛓ 2ec5ec5ded14
✎ 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
2026-06-23
Rubric version
not recorded
Assessed by
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 tests whether the retroviral reverse-transcriptase activity signature observed in ALS is due to a specific endogenous retrovirus (ERV) locus or an entire ERV family, using bioinformatic analysis of post-mortem RNA-seq data across multiple ALS cohorts.

Core claims
  • The ERV locus HML6_3p21.31c is significantly upregulated in ALS post-mortem motor cortex after multiple-testing correction finding
  • HML6_3p21.31c upregulation replicates across independent motor cortex and cerebellum datasets but not in frontal/prefrontal cortex finding
  • HML6_3p21.31c co-expresses with genes enriched for cytokine binding/signalling and responses to EBV, HTLV-1 and HIV Type-1 infection, and this network correlates with ALS disease status finding
  • Neither the HML6 family as a whole nor previously implicated HML2 loci/family show differential expression between ALS and controls finding
  • ERVs upregulated in ALS motor cortex are marginally enriched for SNPs conferring genetic risk of ALS finding
  • HML6_3p21.31c is proximal to and transcriptionally correlated with the HIV-related chemokine receptor genes CCR5 and CCR1 and with LTF finding
  • A bioinformatic pipeline combining the ERVMap protocol, TETranscripts, weighted co-expression network analysis and MAGMA ERV-set analysis can identify locus-specific ERV associations with disease method
  • Specific ERV loci, rather than whole ERV families, are implicated in ALS pathology mechanism
Experimental setups
Assay System Perturbation Readout Platform
bulk RNA-seq (ERVMap pipeline) human post-mortem primary motor cortex (KCL cohort) none (ALS disease vs control) ERV and gene differential expression ERVMap database/protocol
bulk RNA-seq human post-mortem lateral motor cortex (GSE137810) none (ALS disease vs control) ERV differential expression
bulk RNA-seq human post-mortem medial motor cortex (GSE137810) none (ALS disease vs control) ERV differential expression
bulk RNA-seq human post-mortem cerebellum (GSE137810 and GSE67196) none (ALS disease vs control) ERV differential expression (Stouffer's meta-analysis)
bulk RNA-seq human post-mortem frontal/prefrontal cortex (GSE137810 and GSE67196) none (ALS disease vs control) ERV differential expression
cell-type composition deconvolution human post-mortem motor cortex (KCL cohort) none (ALS disease vs control) estimated neuronal proportion BRETIGEA marker database
weighted gene co-expression network analysis (WGCNA) human post-mortem motor cortex (KCL and GSE137810) none (ALS disease vs control) network module co-expression, correlation with disease status, gene function enrichment
ERV-set genetic association analysis human ALS GWAS cohorts (three datasets) none enrichment of ALS-risk SNPs among ERV sets MAGMA
Key results
  • HML6_3p21.31c significantly increased in ALS vs controls in KCL primary motor cortex log2FC=0.691
  • HML6_3p21.31c increased in ALS in GSE137810 lateral motor cortex, similar fold-change to KCL log2FC=0.642
  • Only HML6_3p21.31c remained significant after multiple-testing correction in Stouffer's meta-analysis of KCL and GSE137810 lateral motor cortex Z=5.039
  • HML6_3p21.31c increased in medial motor cortex log2FC=0.645
  • HML6_3p21.31c significantly increased in cerebellum (meta-analysis) but not in frontal cortex regions Z=2.721 (cerebellum), Z=1.22 (frontal, ns)
  • HML6 family and HML2 loci/family show no significant differential expression in ALS
  • HML6_3p21.31c co-expression network enriched for cytokine, EBV and HIV Type-1 related genes, and correlates with disease status r=0.19
  • ERVs upregulated in KCL motor cortex marginally enriched for ALS genetic risk SNPs beta=0.315
Key statistics
  • fold_change log2 fold-change = 0.691, SE = 0.163, p = 2.29 × 10^-5, adjusted p = 0.03 (HML6_3p21.31c, KCL primary motor cortex, ALS vs control)
  • fold_change log2 fold-change = 0.642, SE = 0.220, p = 0.003 (HML6_3p21.31c, GSE137810 lateral motor cortex)
  • pvalue Stouffer's Z = 5.039, p = 4.675 × 10^-7 (Bonferroni threshold 3.546 × 10^-5) (Meta-analysis of HML6_3p21.31c across KCL and GSE137810 lateral motor cortex)
  • fold_change log2 fold-change = 0.645, SE = 0.326, p = 0.04 (HML6_3p21.31c, medial motor cortex)
  • pvalue Stouffer's Z = 2.721, p = 0.006 (HML6_3p21.31c, cerebellum, GSE137810 and GSE67196)
  • correlation r = 0.346, p = 0.0002 (HML6_3p21.31c expression vs CCR5 expression, KCL motor cortex)
  • correlation r = 0.750, p < 2.2 × 10^-16 (HML6_3p21.31c expression vs LTF expression, KCL motor cortex)
  • other beta = 0.315, SD = 0.032, p = 0.039 (MAGMA ERV-set enrichment of ALS-risk SNPs among ERVs upregulated in KCL motor cortex)

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.

The paper used post-mortem RNA-sequencing data to test for differential ERV locus expression in ALS versus controls, applying the ERVMap pipeline in a discovery cohort (KCL primary motor cortex, n=108) and then testing replication across six additional datasets covering multiple brain regions. A Stouffer's meta-analysis combined results across datasets, with Bonferroni correction applied across 3,237 ERV transcripts. Downstream analyses included weighted co-expression network analysis (WGCNA), Pearson correlations with flanking and disease-relevant genes, and MAGMA-based genetic enrichment testing against ALS GWAS summary statistics.

Replicationbiological Sample sizeSample sizes reported per dataset and tissue in Table 1; no formal power calculation described GroupsALS post-mortem donors vs. non-ALS controls across multiple brain regions and datasets Pairingunpaired Randomization/blindingnot stated Dispersionmixed Exact p-valuesyes Effect sizesyes Confidence intervalsno Multiplicity correctionWithin discovery dataset: FDR adjustment (adjusted p-value = 0.03 reported, method not named explicitly, consistent with Benjamini-Hochberg); for Stouffer's meta-analysis: Bonferroni correction for 3,237 ERV transcripts (threshold p < 1.5 × 10⁻⁵)
Statistical tests used
Test Applied to n Assumptions
ERVMap differential expression protocol (likely negative-binomial GLM, specific implementation not named beyond 'ERVMap protocol') ERV locus-level ALS vs. control comparison in KCL primary motor cortex discovery cohort 80 ALS, 28 controls (n=108) not stated
ERVMap differential expression protocol (same as above) Replication in GSE137810 lateral motor cortex, medial motor cortex, frontal cortex, cerebellum; and GSE67196 prefrontal cortex and cerebellum varies by dataset and tissue; see Table 1 not stated
Stouffer's meta-analysis (sample-size weighted Z-score combination of log2 fold-changes) Cross-dataset combination of ERV differential expression results (KCL + GSE137810 motor cortex as primary; GSE137810 + GSE67196 for cerebellum and frontal regions) not stated as a single pooled n; dataset-level n used as weights not stated
Weighted co-expression network analysis (WGCNA); module-trait correlation reported as Pearson's r Co-expression of HML6_3p21.31c with host genes in KCL primary motor cortex and GSE137810 lateral motor cortex (ALS + controls combined) KCL: n=108; GSE137810 lateral MC: n=38 not stated
Pearson's correlation Correlation of HML6_3p21.31c expression with CCR5, CCR1, LTF, TARDBP, optineurin, and 14 ALS-associated genes in KCL primary motor cortex n=108 (KCL cohort) not stated
Linear regression (covariates: sex, age, post-mortem delay, RIN, surrogate variables) for estimated neuron proportions (BRETIGEA) Cell composition sensitivity analysis in KCL cohort n=108 (KCL cohort) not stated
MAGMA gene-set (ERV-set) enrichment analysis Testing whether differentially expressed ERVs were enriched for ALS genetic risk SNPs across three ALS GWAS datasets GWAS sample sizes not stated in this paper not stated
Approaches that could also have been used
  • Stouffer's method (sample-size weighted Z-scores) was used to combine differential expression results across datasets
    Could also: A fixed-effects or random-effects inverse-variance weighted meta-analysis (e.g., as implemented in metafor or meta R packages) could also combine results across datasets — Inverse-variance weighting accounts for heterogeneity in effect-size precision across cohorts rather than weighting solely by sample size; a random-effects model would additionally allow for between-study variance, which is relevant when cohorts differ in tissue sub-region, RIN, or post-mortem delay
  • Pearson's correlation was used throughout for co-expression analyses and gene-level correlations
    Could also: Spearman's rank correlation could also be used for the same comparisons — RNA-seq normalized expression values can be non-normally distributed or contain outlier samples; Spearman's correlation is a nonparametric alternative that is more robust to such deviations and does not assume a linear relationship
  • Within the discovery dataset, an unspecified FDR correction was applied across 1,654 testable ERV transcripts
    Could also: The Benjamini-Hochberg procedure applied to all 3,237 candidate transcripts (not only those passing the read-count filter) would also be a standard approach, and the specific method could be named explicitly — Naming the FDR method and clarifying the denominator of the multiple-testing family (filtered vs. all transcripts) aids reproducibility and interpretation of the adjusted p-value threshold
  • Cell-type composition was estimated using BRETIGEA marker-gene scores and then tested as a covariate in a linear model
    Could also: Reference-based deconvolution methods such as CIBERSORTx or MuSiC (using single-cell RNA-seq reference profiles for brain cell types) could also estimate cell-type proportions — Marker-gene-based scoring and full deconvolution can yield different cell-proportion estimates, particularly in datasets with small control groups; reporting sensitivity analyses across methods would characterise the robustness of the cell-composition adjustment
  • Standard error was reported as the measure of precision around log2 fold-change estimates
    Could also: 95% confidence intervals around the log2 fold-change could also be reported — Confidence intervals convey the same precision information as SE but on a scale that is more directly interpretable for effect-size magnitude and for visual comparison across replication datasets
  • MAGMA gene-set analysis was used to link differentially expressed ERVs to ALS genetic risk
    Could also: LD score regression-based partitioned heritability (S-LDSC) or GARFIELD could also test whether genomic regions overlapping expressed ERVs are enriched for ALS GWAS signal — MAGMA aggregates SNP-level p-values within gene/locus windows, while S-LDSC partitions heritability by functional annotation; the two approaches make complementary assumptions and their concordance would strengthen conclusions about shared genetic architecture
Software: ERVMap · TETranscripts · BRETIGEA · WGCNA (R package, implied by 'weighted co-expression network analysis') · MAGMA · Gene function enrichment tool (references 66–68; specific software not named in extracted text)

What was reproduced

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

Scope — pmid-34253796

Paper: Jones AR, Iacoangeli A, Adey BN, et al. A HML6 endogenous retrovirus on chromosome 3 is upregulated in amyotrophic lateral sclerosis motor cortex. Sci Rep 2021;11:14283. PMID 34253796 · PMCID PMC8275748 · DOI 10.1038/s41598-021-93742-3

Headline result

A single HERV-K(HML6) provirus locus, named HML6_3p21.31c at chr3:46,426,676–46,433,564 (hg38), is upregulated in ALS motor cortex vs control. Reported effect (KCL discovery cohort, motor cortex):

  • log2 fold-change = 0.691, SE = 0.163, p = 2.29×10⁻⁵, adjusted p = 0.03.
  • Cross-cohort meta-analysis (KCL discovery + GSE137810 lateral motor cortex): Stouffer's Z = 5.039, p = 4.675×10⁻⁷.

Pipeline named by the paper

  • ERV quantification: ERVmap curated database + protocol (Tokuyama et al. 2018, PNAS) — a third-party tool applied to the data. Family-level: TEtranscripts.
  • Reference: hg38. Differential expression: DESeq2 with covariates (disease status, sex, age, post-mortem delay, RIN, surrogate variables), IHW + Bonferroni correction.

IN SCOPE (what we attempt — public data + third-party tool)

The discovery (KCL) cohort is NOT public ("available on reasonable request" from the corresponding author → drop_reason data_restricted if taken alone). The paper's public replication data is GSE137810 (NYGC ALS Consortium SuperSeries). Its motor-cortex RNA-seq subseries is GSE124439 (Tam et al. 2019, NYGC/Target ALS), whose raw FASTQ are open access on ENA (BioProject PRJNA512012 / SRP174614).

Reproduction target (1 clear data point): Is the HML6_3p21.31c locus (chr3:46,426,676–46,433,564, hg38) upregulated in ALS vs control lateral motor cortex in the public GSE124439 cohort, using a standard third-party RNA-seq quantification pipeline?

  • Cohort: GSE124439 Motor Cortex (Lateral), ALS Spectrum MND (n=37) vs Non-Neurological Control (n=4). (Paper's GSE137810 lateral MC = 32 ALS/6 ctrl; our public-resolvable counts differ — documented deviation; likely paper QC / one-sample-per-subject / superseries pooling. We use all public lateral-MC ALS+control samples in GSE124439.)
  • Pipeline (run4, faithful to the paper's named tool ERVmap): ENA FASTQ → cutadapt trim → bwa mem to hg38 → ERVmap parse_bam.pl filter → bedtools coverage over ERVmap.bed (HML6_3p21.31c = ERVmap locus id 943) for ERV counts; STAR --quantMode GeneCounts (Ensembl GRCh38.104) for cellular gene counts → DESeq2 size factors → DESeq2 (~condition; + PC sensitivity). Compare direction + significance (and log2FC magnitude vs the public replication value 0.642/p=0.003) of the HML6 locus to the paper.
  • Grading is on direction (upregulated in ALS) and significance, NOT on the exact discovery-cohort log2FC=0.691 (different cohort; the paper itself reports that number only for the restricted KCL data).

OUT OF SCOPE (not attempted; stated honestly)

  1. KCL discovery cohort (80 ALS/28 ctrl) — data_restricted (on request). The exact reported numbers (log2FC=0.691, p=2.29e-5, padj=0.03) cannot be reproduced because the underlying data are not public.
  2. Genome-wide ERVmap over all ~3,220 ERV loci with IHW/Bonferroni multiple-testing — the hard last 20%. We target the single reported locus.
  3. Full covariate model (age, PMD, RIN, surrogate variables) — metadata not all in GEO; we model condition (and sex as a sensitivity). Documented.
  4. GSE67196 (Mayo) replication, GWAS survival, Braineac eQTL — external/secondary.
  5. TEtranscripts family-level analysis.

Why this is a valid reproduction

Per the brief (rule P16): applying an existing third-party tool/pipeline to the paper's own public data is an equally valid reproduction. We test the paper's central biological claim on its own public replication dataset.

Figures / tables: TableFig 1
C1_discovery
Reported
HML6_3p21.31c log2FC=0.691, SE=0.163, p=2.29e-5, adj p=0.03 (KCL discovery, N=80/28, RESTRICTED)
Reproduced
not-attemptable (KCL discovery data 'available on reasonable request' — not public)
partial
C2_replication
Reported
HML6_3p21.31c log2FC=0.642, SE=0.220, p=0.003 (GSE137810 lateral motor cortex, 32 ALS/6 ctrl)
Reproduced
log2FC=3.456, SE=1.173, p=0.00323 (~condition); log2FC=2.348, p=0.0169 (~PC1+PC2+condition) on public GSE124439 lateral MC, 37 ALS/4 ctrl
within tolerance
C3_direction
Reported
HML6_3p21.31c UPREGULATED in ALS motor cortex (central biological claim)
Reproduced
UPREGULATED in ALS: log2FC=+3.456, p=0.00323, ranks #9 by p-value among ERVmap loci
exact
C4_locus
Reported
HML6_3p21.31c = chr3:46426676-46433564 (hg38), an ERVmap locus
Reproduced
ERVmap.bed id 943: '3 46426676 46433564 943 500 +' — exact coordinate + id match
exact

Assessments & scoring basis

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

No assessment has been recorded yet.
🤝
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.

1.8 M
tokens (I/O) · 139.5 M incl. cache
866 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.