Methodology and Statistical Concerns in "RBM4 dictates ESCC cell fate switch from cellular senescence to glutamine-addiction survival through inhibiting LKB1-AMPK-axis" (Signal Transduction and Targeted Therapy, 2023)

Academic fraud report · Geng Detector

Summary

This assessment evaluates the 2023 paper by Chen et al. published in Signal Transduction and Targeted Therapy (DOI: 10.1038/s41392-023-01367-x). The verdict is "questionable": while the mechanistic narrative on RBM4 in esophageal squamous cell carcinoma is internally coherent, several methodological and statistical red flags were identified through text review alone. The most striking issue is a likely typographical (or experimental) error in the CCK-8 protocol, which states a seeding density of 1.5–2 × 10^6 cells per well in a 96-well plate—a value roughly 1000-fold higher than standard practice. Additional concerns include the mechanical use of Dunnett's post-hoc test in rescue experiments where pairwise comparisons (Tukey) would be statistically appropriate, the use of non-standard NE2/NE3 cell lines that cannot be authenticated against ATCC or DSMZ databases, and unusually low inter-animal variability in an in vivo CB-839 efficacy study with only n = 7 mice per group yet P < 0.0001. No image-level duplication was assessed. These findings are indicative of insufficient methodological rigor rather than proven fabrication.

Verdict

Questionable (🟡). The narrative of the paper is internally consistent, but multiple text-based indicators suggest insufficient methodological rigor. Pixel-level image analysis was not performed, so no firm allegation of data fabrication is made.

Key findings

Implausible cell seeding density (likely typo or undisclosed methodology): The CCK-8 cell viability assay (Materials and Methods, page 17) reports seeding at "1.5–2 × 10^6 cells per well" in a 96-well plate, which is approximately 100–1000× higher than conventional densities (typically 1.5–2 × 10^3 to 10^4 cells/well for ESCC lines such as KYSE series). The discrepancy is hard to reconcile with the physical capacity of a 96-well plate.
Statistical test misapplication in rescue experiments: Figures 3c, 3d, 3k, 3l and the Statistics section repeatedly apply "One-way ANOVA with Dunnett multiple comparisons." Dunnett's test compares all groups against a single control; rescue designs (e.g., shRBM4 vs shRBM4+shP27) require pairwise post-hoc testing (e.g., Tukey). The uniform phrasing across panels suggests mechanical copy-pasting of a statistics template.
Non-authenticatable cell lines: The normal esophageal epithelial lines NE2 and NE3 are acknowledged in the STR profiling section (page 16) to lack matches in ATCC and DSMZ databases, leaving their identity formally unverified despite author citations.
Overly tight in vivo variance: Figure 7h–j presents CB-839 efficacy data with n = 7 nude mice per group but extremely small SEM bars and P < 0.0001, an unusually clean separation given the known modest single-agent activity of glutaminase inhibitors in clinical settings.
Image-level analysis not performed: Western blot, SA-β-gal, and colony-formation panels could not be examined at pixel resolution from the supplied PDF (CD2.pdf); no statement on image reuse or splicing is made.

Evidence highlights

Direct quotation from METHODS (Cell viability assay): "Cells were seeded in 96-well plates at a density of 1.5–2 × 10^6 cells per well."
Direct quotation regarding STR: "DNA profiles of NE2 and NE3 are not included in ATCC and DSMZ databases, so the two STR profiles did not match any known cell line and could not be authenticated."
Statistical legend phrasing in Figures 3c/d and 3k/l: "One-way ANOVA with Dunnett multiple comparisons" applied to rescue comparisons.
In vivo experiment: CB-839 treatment vs control, n = 7 per arm, reported significance **P < 0.0001.

Notes

DOI: 10.1038/s41392-023-01367-x
The findings are limited to text-based scrutiny; original uncropped western blots and raw tumor-measurement logs would be required to escalate or refute concerns about data integrity.
The seeding-density issue could conceivably represent a typographical error (× 10^3 intended), but its appearance in the published version of a high-impact journal warrants author clarification.
No definitive evidence of image manipulation, plagiarism, or fabricated data is presented in this report.