Image manipulation and internal inconsistencies in 'Butyrate Suppresses Glucose Metabolism of Colorectal Cancer Cells via GPR109a-AKT Signaling Pathway and Enhances Chemotherapy'

Academic fraud report · Geng Detector

Summary

This report assesses a 2021 Frontiers in Molecular Biosciences paper (DOI: 10.3389/fmolb.2021.634874) and returns a verdict of highly suspicious. Two findings are confirmed as established. First, systematic vertical splicing of Western blots is detected across Figures 1C, 2A/B/D/F, 3D, and 4A: machine forensics found copy-move signals in 4 of 5 images (img-000, img-001, img-003, img-004) with offsets of (0,32) and (56,0), and visual analysis flags abrupt background-texture discontinuities between HCT116 and LoVo lanes without methodological justification. Second, Figure 5A's eight Annexin V-FITC/PI scatter plots show implausibly similar scatter clouds across treatment groups, with img-004.jpg registering 37 copy-move matching blocks at offset (56,0). A third confirmed finding is an internal inconsistency: Figure 3A caption states LC-MS/MS was used for metabolite measurement, while Methods declares CE-connected ESI-TOFMS — distinct analytical platforms. Three additional findings (uniform error bars, an over-regular metabolite heatmap, and suspiciously consistent percentage values) are downgraded because benign explanations (n=3 replicates, strong butyrate effects, Frontiers-style minimal statistics reporting) remain plausible. Overall confidence is high, but final adjudication requires raw blot files, FCS files, and chromatographic data.

Verdict

Highly suspicious. Confirmed image-manipulation signals in Western blots and flow-cytometry scatter plots, plus a cross-section internal methodological contradiction. Independent adjudication requires raw data from the authors.

Key findings

Established: Systematic vertical splicing of Western blots across multiple main figures (Figures 1C, 2A/B/D/F, 3D, 4A). Both visual analysis (abrupt background-texture discontinuities at HCT116/LoVo junctions) and machine forensics (copy-move signals in 4 of 5 images at offsets (0,32) and (56,0)) independently converge. The paper's Methods describes a single SDS-PAGE workflow without declaring any cross-gel assembly rationale.
Established: Figure 5A's eight Annexin V-FITC/PI scatter panels (HCT116 ± butyrate ± 5-FU ± SC79) display implausibly similar scatter-cloud morphologies given the highly divergent apoptotic stimuli; img-004.jpg registers 37 copy-move matching blocks at offset (56,0).
Established: Internal methodological contradiction — Figure 3A caption states "liquid chromatograph-mass spectrometer/mass spectrometer (LC-MS/MS)", while Methods states metabolite analysis used "a capillary electrophoresis (CE)-connected ESI-TOFMS system". LC-MS/MS and CE-TOFMS are distinct separation-detection platforms; no benign reinterpretation is available.
Downgraded (benign explanations plausible): Visually uniform error bars in Figures 3E, 4B, 4C (n=3 replicates may legitimately yield similar SDs; raw SD values not reported).
Downgraded (low-confidence visual cue; Methods declares measurement platform): Figure 3A metabolite heatmap appears over-regular, with reported reduction percentages clustering in the 58–70% range.
Downgraded (Frontiers-style minimal statistics reporting; strong biological effect sizes plausible): Multiple percentage values in the Results appear "textbook-perfect" and statistical reporting is limited to p<0.05 without full test statistics.

Evidence highlights

[img-000.jpg] copy-move detection at offset (0,32): 54 matching blocks (image/copy_move)
[img-003.jpg] copy-move detection at offset (0,32): 14 matching blocks (image/copy_move)
[img-004.jpg] copy-move detection at offset (56,0): 37 matching blocks (image/copy_move)
[img-001.jpg] and [img-004.jpg] also show copy-move signals
Noise-variance inconsistencies flagged across multiple images ([img-000], [img-002], [img-003], [img-005], [img-007])
Figure 3A caption verbatim: "were measured by liquid chromatograph-mass spectrometer/mass spectrometer (LC-MS/MS)"
Methods verbatim: "analyzed using a capillary electrophoresis (CE)-connected ESI-TOFMS system"
Methods verbatim: "The error bars showed standard deviations from independent triplicates"
Reported metabolite reductions: lactate 60%, R5P 70%, acetyl-CoA 58%, NADPH 60% (Supplementary Figure S4)
DOI: 10.3389/fmolb.2021.634874

Notes

The Bayesian synthesis reports a posterior ≥99.9% (BF ≈ 3.51×10⁶) under a fixed 5% prior with conditional-independence assumptions; this figure is likely an overestimate due to correlated imaging evidence and should not be treated as a verdict.
Five forensic lines were excluded during review because cited numeric statistics (e.g., MAD=0.0629, χ²=51.2, χ²=503.8, 12/91=0.13) do not appear in the source text — these are machine-computed outputs, not author-reported values.
Recommended follow-up: request unedited Western blot files, FCS files for Figure 5A, and original chromatograms (CE-TOFMS or LC-MS/MS, with clarification of which platform was actually used), and request underlying replicate-level data enabling independent statistical re-analysis.
Severity scale used by reviewers: 🔴 critical, 🟠 major, 🟡 minor. Findings 1, 2 = 🔴; Finding 3 = 🟠; Findings 4–6 = 🟡 → ⚠️ after benign-explanation review.