Image and statistical irregularities in Chem. Commun. 2015, 51, 11665–11668 (DOI: 10.1039/c5cc03662c): The synthesis and evaluation of near-infrared probes with barbituric acid acceptors for in vivo detection of amyloid plaques

Academic fraud report · Geng Detector

Summary

Targeted review of a 2015 Chem. Commun. paper (DOI: 10.1039/c5cc03662c) by Zhou, Fu, Feng, Cui, Dai, and Liu on barbituric-acid NIR probes for amyloid plaque imaging. Overall assessment: highly suspicious. Three findings are flagged: (1) In Figure 2, the fluorescence emission spectra of probe BBTOM-3 bound to Aβ1-40 aggregates and Aβ1-42 aggregates appear almost perfectly overlapping despite a roughly two-fold difference in reported binding affinities (Ki = 23.9 nM vs 13.7 nM), suggesting data reuse or post-hoc curve tracing. (2) In Figure 4, in vivo brain kinetics for Tg and WT groups show error bars that are abnormally uniform in length and highly symmetric across the 5–15 min signal-change window, inconsistent with the variability expected from n = 3 animals. (3) Figure 3 histological panels show unusually uniform backgrounds across panels, raising concerns about excessive processing. Evidence is circumstantial; raw spectra, ROI data, and original images must be examined before misconduct can be confirmed.

Verdict

🟠 Highly suspicious. Multiple figure-level irregularities were identified that warrant a request for raw data and original images. The observations are consistent with potential data reuse and image processing/manipulation but do not by themselves constitute proof of misconduct.

Key findings

Figure 2 (Page 11666) – Suspicious spectral overlap: Emission spectra for BBTOM-3 bound to Aβ1-40 vs Aβ1-42 aggregates (red vs purple dashed lines) show near-identical peak shape, full width at half-maximum, and decay slope, despite markedly different binding affinities reported in Table 1 (Ki = 23.9 nM for Aβ1-40; Ki = 13.7 nM for Aβ1-42). Lack of natural noise and inter-spectrum variation raises suspicion of data reuse or manual curve redrawing.
Figure 4 (Page 11667) – Implausible error-bar uniformity: In vivo NIR brain-imaging kinetics for Tg (red) and WT (black) groups display error bars of nearly constant length and unusually high symmetry across multiple time points, particularly during 5–15 min. With the reported n = 3, such invariance in error-bar magnitude is statistically atypical for in vivo imaging.
Figure 3 (Page 11667) – Unnaturally uniform tissue backgrounds: Fluorescence-stained AD transgenic mouse and AD patient brain sections exhibit extremely clean backgrounds. While background subtraction can produce this effect, the suspiciously uniform grayscale across panels A–F raises concerns about over-processing or possible local image replacement.

Evidence highlights

DOI: 10.1039/c5cc03662c (Chem. Commun., 2015, 51, 11665–11668).
Quantitative anchor: Ki(Aβ1-40) = 23.9 nM vs Ki(Aβ1-42) = 13.7 nM from Table 1, contrasted against near-identical spectral traces in Figure 2.
Sample size explicitly stated as n = 3 in the in vivo imaging experiment, yet error bars in Figure 4 appear unnaturally invariant.
Figures cited: Figure 2 (Page 11666), Figure 3 panels A–F (Page 11667), Figure 4 brain-kinetics panel (Page 11667).

Notes

All three findings are based on visual/statistical heuristics applied to the published figures; conclusions remain preliminary without access to the underlying raw spectra, per-animal ROI fluorescence values, and unprocessed histology images.
Follow-up actions are recommended: (1) request raw spectral and ROI imaging data from the authors; (2) raise the concern on PubPeer; (3) report to the journal editorial office; (4) if warranted, notify the authors' institutional research integrity committee.
This automated assessment is intended for academic discussion only; any final determination of misconduct requires a formal institutional investigation. False positives and false negatives are possible.