Gut-derived Lactobacillus from exceptional responders mitigates chemoradiotherapy-induced intestinal injury through methionine-driven epigene
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.
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.
- ✓Reported values were directly comparable
- ✓The central claim held under reproduction
- 🟡Could not use the authors’ exact input data
- 🔴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
- 🟡Overall, the reproduction showed a material discrepancy
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 from the authors' own shipped figure-source CSVs (github.com/yulugithub/iMeta @ b0c1789), NOT from the harvested GEO accession. NOTE: the RU's data accession geo:GSE124880 is the wrong dataset (it is Xu et al. 2019 / PMID 31618654, only cited here as a reference atlas) - so per BRIEF rule P16 we reproduced against the paper's own shipped data. 10 of 11 numeric claims reproduce (7 exact, 2 within-tol, plus 2 consistency checks where the shipped CSV equals the reported number); genuine recomputations include the random-forest AUC argmax (0.854 @ 15 feat), differential-metabolite up/down counts (52/65), KEGG pathway counts, the metB pan-genome percentages (S6B 41.47%, S6C 20.83% exact), and one rank-sum p-value (Fig 7N 0.256 vs 0.2604). ONE MISMATCH worth a human look: Fig 1B alpha-diversity p=0.5335 is not reproducible from the shipped ACE values by any standard test (rank-sum/Welch/Student all ~0.81-0.83); conclusion (n.s.) still holds, likely a different index/test not shipped. Compute was trivial (counts + 2 rank-sum tests on <=61 numbers) = the scripting class the brief allows on the control plane; there was no heavy compute to send to «our HPC» and no raw/large data was downloaded to «host» («our HPC» VPN also needed human 2FA which did not complete this session). NOT attempted (hard ~20%, inputs not shipped): full GFF-based gene comparison (no genome accession), full clusterProfiler/enrichKEGG & GSVA recompute (gene lists/expression matrices not shipped), PERMANOVA beta-diversity p-values (distance matrix not shipped), LEfSe/LDA scores and Fig 7I correlations (precomputed outputs only), and all wet-lab results.
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.
-
v1 current initial assessment Score 89assessed: 2026-06-15 ⛓ fa7cf54b4dea
✎ 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
- 2026-06-15
- Rubric version
- v1.0
- Assessed by
-
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15no 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: opusThe study tests whether the gut microbiome—specifically a high abundance of Lactobacillus—can mitigate acute chemoradiotherapy-induced intestinal injury (ACRIII) in colorectal/rectal cancer patients, and seeks to identify patient-derived Lactobacillus strains and the mechanism by which they protect the intestine.
- ★ High pretreatment gut abundance of Lactobacillus (and Bifidobacterium) is strongly associated with the absence of ACRIII in rectal cancer patients undergoing neoadjuvant chemoradiotherapy. finding
- ★ Ten novel patient-derived Lactobacillus strains were isolated from exceptional responders (complete remission without ACRIII), with L. rhamnosus DY801 emerging as the most effective at mitigating ACRIII. resource
- ★ L. rhamnosus DY801 synthesizes methionine via the metB gene, modulating methionine metabolism in host gut lymphoid tissue inducer (Lti) cells. mechanism
- ★ Microbial-derived methionine increases intracellular S-adenosylmethionine (SAM) and enhances histone H3K4 trimethylation (H3K4me3) in Lti cells, suppressing pro-inflammatory IL-17A and IL-22 and reducing ACRIII severity. mechanism
- ★ FMT in a mouse model of chemoradiotherapy-induced injury partially alleviates diarrhea, weight loss, mortality, intestinal permeability, and villus damage. finding
- A Random Forest classifier on genus-level metagenomic features identifies Lactobacillus as the most important genus distinguishing ACRIII from non-ACRIII patients. method
- ★ DY801 mitigates ACRIII without diminishing the therapeutic efficacy of chemoradiotherapy (strains sensitize tumor cells while protecting normal intestinal epithelium). finding
- Isolated strains exhibit superior acid resistance and bile salt tolerance versus ATCC strains, supporting clinical applicability and safety. finding
| Assay | System | Perturbation | Readout | Platform |
|---|---|---|---|---|
| Shotgun metagenomic sequencing + bioinformatics (alpha/beta diversity, LEfSe/LDA, KEGG, Random Forest) | baseline fecal samples from 61 rectal cancer patients undergoing neoadjuvant chemoradiotherapy | none (observational, ACRIII vs non-ACRIII) | microbial taxa abundance, KEGG pathway enrichment, classifier AUC | — |
| Fecal microbiota transplantation (FMT) experiment + metagenomics of cecal contents | mouse model (whole-abdomen irradiation + intraperitoneal 5-fluorouracil) | FMT vs RCT (radiochemotherapy) vs Control | diarrhea/fecal output, body weight, survival/mortality, villus length, microbiota composition | — |
| Intestinal permeability assay | mice (RCT model) | RCT vs FMT vs Control | FITC-dextran permeability | fluorescein isothiocyanate–labeled dextran (FITC-dextran) |
| Histopathology (H&E staining) | mouse intestinal segments (jejunum, ileum) and vital organs | RCT/FMT; acute toxicity of strains | villus length, pathological damage, organ indices | hematoxylin and eosin staining |
| Acid/bile salt resistance survival assay | 10 Lactobacillus isolates vs ATCC strains in artificial gastric/intestinal fluid | varying pH (2,3,4), bile salt | survival rate | — |
| Bacterial adhesion assay (CFU) | primary intestinal epithelial cells (CK-18 identified) | Lactobacillus strains vs ATCC | CFU adhering to cells | — |
| Cell viability coculture assay | primary intestinal epithelial cells and MC38 murine colon cancer cells | heat-pasteurized (inactivated) bacteria and bacterial supernatants | cell survival/viability | — |
| Comet assay and IC50 / fluorouracil chemosensitivity assay | irradiated primary intestinal epithelial cells and MC38 cells | bacterial supernatants from Lactobacillus strains; radiation; fluorouracil | % DNA in comet tail (radiation resistance), IC50 of fluorouracil | — |
- ▲ Non-ACRIII patients showed elevated Lactobacillaceae, Lactobacillus, Limosilactobacillus, Bifidobacteriaceae and Bifidobacterium (LEfSe/LDA).
- – Random Forest classifier achieved highest AUC with 15 top features. AUC = 0.854
- – 86 differential KEGG pathways identified (36 enriched in ACRIII, 50 in non-ACRIII); cysteine and methionine metabolism enriched in non-ACRIII. 86 pathways
- ▼ RCT mice developed severe diarrhea/weight loss and 100% mortality by Days 5-6; FMT improved weight loss and reduced mortality. 100% mortality in RCT
- ▲ DY801 showed the highest survival rate at pH 2 among isolates. ~38% at pH 2
- – All isolated strains showed high bile salt tolerance with no significant difference from ATCC controls. >90%
- ▼ RCT reduced jejunal villus length (partially restored by FMT) and ileal villus length (completely restored by FMT).
- – CYQ09, DY801, and DY802 enhanced radiation resistance of primary intestinal epithelial cells while sensitizing MC38 tumor cells to radiation/chemotherapy.
- other AUC = 0.854 (Random Forest classifier with 15 most important genus features)
- pvalue p = 0.5335 (alpha diversity (Ace) between ACRIII and non-ACRIII groups, not significant)
- count 86 KEGG pathways (36 ACRIII-enriched, 50 non-ACRIII) (differential KEGG pathway analysis of metagenomic data)
- other PC1: 20.40%, PC2: 13.03% (PCoA beta diversity in patient cohort)
- other PC1: 91.54%, PC2: 4.32% (beta diversity of mouse cecal metagenomes (Control/RCT/FMT))
- count 100% mortality (RCT-group mice euthanized by Days 5-6 due to acute diarrhea)
- other ~38% (DY801 survival rate at pH 2 in artificial gastric fluid)
- count 61 patients; 10 strains isolated (rectal cancer patient cohort and number of novel Lactobacillus strains isolated)
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 study combined clinical metagenomics (n=61 rectal cancer patients), a mouse chemoradiotherapy model with FMT, and a multi-criterion in vitro/in vivo strain-screening pipeline. Gut microbiota associations with ACRIII were assessed via alpha/beta diversity, LEfSe, and a Random Forest classifier; downstream mechanistic experiments used standard significance thresholds (*p<0.05, **p<0.01, ***p<0.001) with one exact p-value reported for alpha diversity. In vitro and in vivo comparisons evaluated Lactobacillus strains against ATCC reference strains across eleven scored criteria.
| Test | Applied to | n | Assumptions |
|---|---|---|---|
| Unspecified test for alpha diversity (Ace index) between ACRIII and non-ACRIII groups | Figure 1B — clinical cohort alpha diversity comparison | 61 patients | not stated |
| Principal Coordinate Analysis (PCoA) — ordination; permutation significance test (e.g., PERMANOVA) not explicitly named | Figure 1C — beta diversity between ACRIII and non-ACRIII; Figure S1K — mouse cecal metagenomics | 61 patients (clinical); not stated (mouse) | not stated |
| Linear Discriminant Analysis Effect Size (LEfSe) with LDA scores | Figures 1E, 1F — differential bacterial taxa between ACRIII and non-ACRIII | 61 patients | not stated |
| Random Forest classifier (machine learning; mean decrease in accuracy for feature importance; AUC reported) | Figures 1J–1L — genus-level classification of ACRIII vs. non-ACRIII; AUC = 0.854 with 15 features | 61 patients | not stated |
| Unspecified statistical tests for pairwise group comparisons (reported as NS/*/ **/***) | Figures 2B–2H, S1C–S1I — strain survival, adhesion, cell viability, comet assay, IC50 comparisons against ATCC controls; mouse weight, survival, FITC-dextran permeability, villus length | two or three independent experiments (stated in figure legends); exact n not stated | not stated |
| KEGG pathway differential enrichment analysis (method not explicitly named) | Figures 1G–1I — 86 KEGG pathways compared between ACRIII and non-ACRIII metagenomic data | 61 patients | not stated |
-
Alpha diversity was summarized using the Ace index alone, and the test used to compare groups was not named↳ Could also: Multiple complementary indices (e.g., Shannon entropy, Chao1, Simpson's index) could be reported together, and the specific test (e.g., Wilcoxon rank-sum, Kruskal-Wallis) could be named — Different diversity indices capture richness vs. evenness differently; naming the test allows readers to assess assumptions (e.g., normality) and reproduce the p-value
-
Beta diversity was visualized with PCoA; no permutation-based significance test result (e.g., PERMANOVA R² and p-value) was reported↳ Could also: PERMANOVA (adonis in R/vegan) or ANOSIM could be run alongside PCoA to provide a formal test statistic and p-value for between-group separation — Visual clustering in PCoA can be misleading with unequal group dispersions; a significance test with effect size (R²) quantifies the proportion of variance explained by group membership
-
LEfSe was used to identify differentially abundant microbial taxa between ACRIII and non-ACRIII groups↳ Could also: Negative-binomial models (e.g., DESeq2, ANCOM-BC, or MaAsLin2) could also be used for differential abundance testing on metagenomics count data — These approaches explicitly model count overdispersion and apply FDR correction across all taxa simultaneously, providing adjusted p-values and log2 fold-changes with confidence intervals
-
Multiple pairwise comparisons between individual Lactobacillus strains and ATCC controls (across 11 criteria in Figure 2) were performed with unnamed tests and significance stars↳ Could also: A one-way ANOVA (or Kruskal-Wallis) with Dunnett's post-hoc test (control vs. each strain) could also be applied to this family of comparisons — Dunnett's test is designed specifically for multiple treatment-vs.-control comparisons and controls the family-wise error rate, which is relevant when 10+ strains are each compared to the same reference
-
The Random Forest classifier was evaluated primarily by AUC; confidence intervals were noted but model calibration and cross-validation scheme were not detailed in the provided text↳ Could also: Repeated k-fold cross-validation or bootstrap resampling with calibration curves could also be reported alongside AUC — With n=61, overfitting risk in Random Forest is non-trivial; repeated CV and calibration plots convey whether the model's probability estimates are reliable, not just its rank discrimination
-
Continuous outcomes in cell-based assays (viability, comet tail DNA %, IC50) were compared between groups using unnamed tests with threshold-based p-value reporting↳ Could also: Reporting effect sizes (e.g., Cohen's d or fold-change with 95% CI) alongside named tests (e.g., one-way ANOVA with Dunnett's or Tukey's HSD) would complement the significance stars — Effect sizes communicate practical magnitude independently of sample size, and naming the test allows assessment of whether distributional assumptions (normality, homoscedasticity) were appropriate for the small n per group
Citation network
Where this publication sits in the reproducibility-weighted citation graph — what it is built on, and what is built on it. Citation data from OpenAlex.
No assessed neighbours yet — the network grows as more papers are assessed.
What was reproduced
The exact results taken into scope, with each reported value next to the value our attempt produced.
Scope — pmid-40469520
Paper: Gut-derived Lactobacillus from exceptional responders mitigates chemoradiotherapy-induced intestinal injury through methionine-driven epigenetics. imeta 2025, DOI 10.1002/imt2.70043 · PMID 40469520 · PMCID PMC12130557.
Code: https://github.com/yulugithub/iMeta (authors' own; default branch main,
last push 2025-03-28). Public, not archived, no license file.
Data-provenance correction (important)
The RU's harvested data accession geo:GSE124880 is NOT this paper's deposited
data. GSE124880 is "Comprehensive transcriptional atlas of intestinal immune
cells … α-CGRP" (Xu et al. 2019, PMID 31618654). This paper cites GSE124880 only
as a reference single-cell atlas for cell-type annotation in the spatial-
transcriptomics GSVA step (see Scripts/GSVA.R → Cell_annotation.csv). The
paper's own pipeline-derived results are shipped directly as per-figure source-
data CSVs in the GitHub repo. Reproduction therefore runs against the repo's
shipped CSVs, not against GSE124880. (Per BRIEF rule P16, applying analysis to the
paper's own shipped data is fully valid.)
What the repo ships
- Per-figure source-data CSVs (Fig 1,4,5,7,S1,S6,S7,S8) — the already-computed values behind each panel.
- 6 R scripts (
Scripts/, duplicated in figure folders):Correlation.R,GSVA.R,Gene_comparison_summary.R,clusterProfiler.R,clusterProfiler_DEGs_between_DY801_and_ATCC.R,ggplot2_fill_bar.R.
IN SCOPE — pipeline-derived, recomputable from shipped data
| Result | Panel | Pipeline | How reproduced |
|---|---|---|---|
| Alpha diversity ACRIII vs non-ACRIII, p=0.5335 | Fig 1B | rank-sum test on ACE | recompute Wilcoxon/MWU from per-sample ACE |
| PCoA variance PC1 20.40% / PC2 13.03% | Fig 1C | PCoA | shipped file == reported (consistency) |
| Random-forest AUC max 0.854 at 15 features | Fig 1L | RF + ROC | recompute argmax over shipped AUC-vs-#features |
| Gene comparison 63 unique / 62> / 46< | Fig 4A | GFF CDS comparison | consistency vs shipped (full GFF recompute out-of-scope) |
| 19 enriched KEGG pathways (DY801 DEGs) | Fig 4B | clusterProfiler enrichKEGG(lrh) | count shipped pathways (full enrichKEGG out-of-scope) |
| Metabolites 52 up / 65 down (117 total) | Fig 4D | OPLS-DA / threshold | recompute up/down counts from volcano table |
| 8 KEGG pathways Q<0.05 (metabolomic) | Fig 4E | KEGG enrichment | recompute count of Q<0.05 rows |
| Alpha diversity 3rd cohort, p=0.2604 | Fig 7N | rank-sum on ACE | recompute Wilcoxon/MWU |
| 58% Lactobacillus harbor metB | Fig S6A | pan-genome count | recompute percentage |
| Bifidobacterium 41.47% of single-metB | Fig S6B | pan-genome count | recompute percentage |
| Lactobacillus 20.83%, rank 2 of ≥2-metB | Fig S6C | pan-genome count | recompute percentage + rank |
OUT OF SCOPE / not attempted (and why)
- Full GFF-based gene comparison (Fig 4A inputs) — DY801 & ATCC_53103 genome
GFFs are not shipped and no genome accession is given in the paper; only the
summary counts are shipped. We verify the shipped summary == paper, not the GFF
parse. (drop-class for that sub-step:
no_data_accession.) - Full clusterProfiler/enrichKEGG recompute (Fig 4B, 5G) and GSVA (Fig 5D) —
upstream inputs (
gene.txt,DEG1.txt,Cell_annotation.csv, the expression matrix) are not shipped; only outputs are. We verify output counts only. - PERMANOVA beta-diversity p-values (Fig 7O p=0.7690 etc.) — require the full
distance matrix / OTU table; the shipped
PCoA.csvholds only variance proportions. Not recomputable. - LEfSe/LDA scores (Fig 1E/F), correlation coefficients (Fig 7I) — shipped as precomputed outputs; the scripts only plot them. No recompute possible.
- All wet-lab results (16S strain ID, organoids, CUT&Tag wet steps, animal phenotypes, flow cytometry %, GeoMx wet steps) — not computational pipeline outputs; not attempted.
Compute placement
Recomputations are lightweight (counts, perce
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.
Reproduction is strong: 10 of 11 numeric claims match (7 exact, 2 within-tol, plus 2 consistency checks) using the authors' own shipped figure-source CSVs, since the harvested accession GSE124880 is the wrong dataset. The one substantive issue is on the authors'/data side: Fig 1B's reported alpha-diversity p=0.5335 is not derivable from the deposited ACE values by any standard test (all ~0.81–0.83), most likely an undisclosed index/test — but the non-significant conclusion is preserved, so no central claim is overturned. Minor extras: a Fig 4D label vs caption inconsistency and reliance on consistency checks because raw genomes/gene lists/distance matrices were never shipped. Overall yellow — solid with explainable, non-fatal deviations.
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-4-8Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.