Image integrity review: RTA408 alleviates retinal ganglion cells damage in mouse glaucoma by inhibiting excessive autophagy (PLOS ONE, DOI: 10.1371/journal.pone.0313448)

Academic fraud report · Geng Detector

Summary

This automated forensic review of the 2024 PLOS ONE paper by Qian et al. (DOI: 10.1371/journal.pone.0313448) returns a verdict of 'suspicious / inconclusive.' Three classes of concern were flagged: (1) possible splice lines and copy-move matches across Western blot panels in Figures 2C, 3B, 6A, and 7A, with detectable background-step discontinuities between lanes and inconsistent β-actin noise patterns; (2) apparent reuse of retinal whole-mount images between Figure 1D and Figure 4A, and between Figure 2A and Figure 5A; (3) unusually tight, near-linear scatter distributions in the quantification plots of Figures 2D, 3B, and 6B. Each finding has plausible benign explanations — layout cropping, serial sections from the same control/COH cohort, and instrument step-size or default jitter in GraphPad — so the report stops short of a definitive misconduct judgment. Automated Bayesian synthesis yields a posterior ≥99.9% with BF≈7.91e100, but this is derived under a fixed 5% prior with assumed conditional independence, which likely overstates joint strength. Confidence is moderate; final determination requires the unprocessed full-membrane images in S1 File and the underlying numerical data.

Verdict

Suspicious — inconclusive. Multiple image-integrity signals were detected, but each admits benign technical explanations, and the authors claim full unprocessed blots are deposited as supplementary material. A formal misconduct finding cannot be made on the present evidence.

Key findings

Possible Western blot splicing (Figures 2C, 3B, 6A, 7A): Visible vertical background-step discontinuities between lanes; β-actin loading-control noise patterns differ across lanes; forensic copy-move detection flags matching blocks with non-zero offsets.
Suspected image reuse across panels (Figure 1D ↔ Figure 4A; Figure 2A ↔ Figure 5A): Control and COH retinal whole-mount images appear highly similar across different experimental sections.
Anomalously tight scatter distributions (Figures 2D, 3B right panel, 6B): Data points in several quantification plots are arranged in near-horizontal lines with minimal biological variance, despite the methods declaring mean ± SEM.
Each concern has a plausible benign alternative: cropping/relayout for figures 2C/3B/6A/7A, serial sections from a shared cohort for 1D/4A and 2A/5A, and instrument step-size or default GraphPad jitter for the scatter plots.

Evidence highlights

DOI: 10.1371/journal.pone.0313448
Copy-move detections reported by the forensic pipeline (illustrative subset):
img-039.png — offset (-96, 112), 8 matching block pairs, logLR = 3
img-027.png — offset (-160, 208), 8 matching block pairs, logLR = 3
img-025.jpg — offset (32, 0), 48 matching block pairs, logLR = 3
img-015.png — offset (-96, 112), 8 matching block pairs, logLR = 3
img-014.jpg — offset (48, 0), 34 matching block pairs, logLR = 3
Methods text confirms Western blot workflow; figure legends reference protein-level readouts (e.g., BAX, BCL-2) and retinal flat-mount imaging of RGCs.
Authors state that unprocessed original blots are provided in S1 File (legibility of supplementary material not independently verified in this review).

Notes

The automated Bayesian summary (posterior ≥99.9%, BF ≈ 7.91 × 10¹⁰⁰, CI [100%, 100%]) is generated under a fixed 5% prior with conditional-independence assumptions, which can inflate joint evidence when multiple forensic signals originate from the same figure. Treat the numerical headline as an upper bound, not a calibrated probability.
Recommended follow-up: (i) request and inspect S1 File unprocessed full-membrane images; (ii) request raw numerical data behind the scatter plots and the section/series identifiers linking the 1D/4A and 2A/5A whole-mount panels; (iii) raise the concerns on PubPeer for community verification.
This report is AI-assisted and intended for academic discussion only; final adjudication of misconduct requires institutional investigation.