Geng Integrity Assessment Report: 'Effect of Silencing HCCR-2 Gene on Bcl-2 and Bax Expression, Cell Proliferation and Apoptosis in Pancreatic Cancer Cells'

Academic fraud report · Geng Detector

Summary

This report evaluates a 2012 paper in the Chinese Journal of Clinicians (Electronic Edition) by Jiang Lin et al. (DOI: 10.3877/cma.j.issn.1674-0785.2012.22.110) for data integrity. The verdict is highly suspicious, based primarily on severe statistical incoherence in the reported quantitative data. Back-calculation of independent-sample t tests from the reported means and standard deviations (n=3) yields t≈1.22, P≈0.27 for MTT results (antisense 2.38±1.03 vs nonsense 3.78±1.69) and t≈1.32, P≈0.24 for HCCR-2 mRNA (antisense 0.29±0.16 vs nonsense 0.55±0.31), yet the paper claims P<0.01 throughout. Standard deviations cluster suspiciously at 43–60% of group means across tables, and MTT A490 values for the normal-culture control group show a non-monotonic, biologically implausible decline (48 h: 3.83±1.43 → 72 h: 3.05±1.85), violating basic expectations of pancreatic cancer cell proliferation. Image analysis was not possible due to corrupted PDF text. Overall confidence in the statistical findings is high; confidence in image-related conclusions is low pending original data.

Verdict

Highly suspicious. The paper claims uniformly extreme statistical significance (P < 0.01) across all core experiments, but reverse t-test calculations from the reported means and SDs show that such P values are mathematically unattainable with the given sample size and variance. Additional anomalies in the SD distribution and in longitudinal MTT values further suggest fabricated or invented numerical data. Image-based verification was not possible due to PDF text/image extraction limitations.

Key findings

Statistical incoherence (P-value fabrication): Back-calculated independent-sample t tests for both MTT (antisense 2.38±1.03 vs nonsense 3.78±1.69, n=3) and HCCR-2 mRNA (antisense 0.29±0.16 vs nonsense 0.55±0.31, n=3) yield P ≈ 0.27 and P ≈ 0.24, contradicting the reported P < 0.01.
Suspiciously uniform SDs: Standard deviations across Tables 1–3 cluster tightly between ~43% and ~60% of their respective means (e.g., 2.38±1.03 ≈ 43%; 3.78±1.69 ≈ 44%; 0.47±0.28 ≈ 59%; 0.93±0.48 ≈ 51%), inconsistent with natural biological variability.
Implausible MTT time course: In the normal-culture control, the A490 value drops from 3.83±1.43 at 48 h to 3.05±1.85 at 72 h, and the nonsense-control group shows no monotonic increase from 24 h to 72 h (3.78±1.69 → 3.68±1.37 → 3.88±1.45), contradicting expected pancreatic cancer cell proliferation.
Excessively large SDs in MTT: SDs of 1.03 and 1.69 on A490 values of 2.38 and 3.78 are implausibly high for biological replicates and would typically be flagged as unreliable in real MTT assays.
Image analysis not possible: PDF text was corrupted; flow cytometry, fluorescence, and Western blot images in Figures 1–4 could not be examined for duplication or splicing.

Evidence highlights

DOI: 10.3877/cma.j.issn.1674-0785.2012.22.110
MTT (Table 3 / abstract): antisense 2.38 ± 1.03 vs nonsense 3.78 ± 1.69 (n = 3); back-calculated t ≈ 1.22, P ≈ 0.27 vs reported P < 0.01.
HCCR-2 mRNA (Table 1): antisense 0.29 ± 0.16 vs nonsense 0.55 ± 0.31 (n = 3); back-calculated t ≈ 1.32, P ≈ 0.24 vs reported P < 0.01.
SD-to-mean ratio systematically in the 43–60% band across all numerical tables.
Normal-culture control A490: 3.83 ± 1.43 (48 h) → 3.05 ± 1.85 (72 h), a ~20% non-monotonic decrease.

Notes

All quantitative claims rely on back-calculation from reported mean ± SD with n = 3; if the true sample size was larger, the t-test conclusions would be partially mitigated, but this would itself contradict the paper's reported methods.
The biological-plausibility issues (non-proliferating tumor cells) and the uniformity of SD ratios are soft signals consistent with synthetic data generation but are not, individually, definitive.
Figures 1–4 (flow cytometry, fluorescence microscopy, Western blots) remain unexamined; an institutional investigation should request original FCS files, uncropped blot images, and raw plate reader exports.
Confidence: high for the statistical anomalies; low for any image-related conclusions due to technical limitations of the source PDF.
This assessment does not constitute a formal finding of misconduct; final determination requires institutional review.