Image Integrity Concerns in 'Antisense oligonucleotide targeting TARDBP-EGFR splicing axis inhibits progression of oral squamous cell carcinoma through ABCA1-regulated cholesterol efflux' (DOI: 10.1038/s41368-025-00402-7)

Academic fraud report · Geng Detector

Summary

This report assesses the Geng academic fraud detection findings against an article published in International Journal of Oral Science (DOI: 10.1038/s41368-025-00402-7). The assessment reaches a verdict of highly suspicious. The primary issue is systematic Western blot image splicing observed across multiple panels (Figures 2d, 2g, 3a, 3c, 5a, 6a, 7e, and 7f), including vertical white-line artifacts and abrupt background discontinuities, with the GAPDH loading control itself appearing cut and re-assembled. A secondary concern is inconsistency in sample size reporting: the abstract describes 67 OSCC cases for sQTL analysis, while Figure 1a is annotated as '81 pairs' of differentially expressed transcripts, with no reconciliation offered in the figure legend. These findings raise concerns about data integrity and internal consistency, but conclusions remain tentative pending review of original uncropped blot images and clarification of cohort definitions by the authors and the journal.

Verdict

Highly suspicious. Multiple Western blot panels show apparent splicing artifacts, and the manuscript contains a notable inconsistency in reported sample sizes between the abstract and Figure 1.

Key findings

Western blot image splicing across multiple figures: Panels 2d, 2g, 3a, 3c, 5a, 6a, 7e, and 7f exhibit visible vertical white-line artifacts and abrupt background/texture discontinuities between lanes; the GAPDH loading-control bands also appear cleanly severed along these lines.
Failure to mark spliced regions: None of the affected blots appear to declare splicing boundaries in figure legends or annotations, contrary to standard journal practice.
Sample-size inconsistency between abstract and Figure 1a: The abstract states 67 OSCC cases (and 67 normal epithelial samples for sQTL), whereas Figure 1a is annotated as involving 81 paired OSCC/adjacent-normal samples; the figure sits under the sQTL-figure title, with no legend-level clarification.

Evidence highlights

Figure 2d and 2g: Vertical white lines between lanes of different cell lines/conditions; GAPDH bands cleanly bisected.
Figure 3a and 3c: Unnatural vertical gaps between Input lanes and treatment lanes; background-tone discontinuity between TARDBP and GAPDH regions.
Figure 5a and 6a: Systematic vertical demarcations across time-point and treatment lanes with background discontinuities.
Figure 7e and 7f: Background noise pattern of Vimentin and GAPDH bands discontinuous across vertical lines.
Abstract text: "We collected surgically resected cancer and adjacent normal epithelial tissue samples from 67 OSCC cases ... A genome-wide sQTL analysis was performed using the RNA sequencing data from 67 normal oral epithelial tissue samples."
Figure 1 / Figure 1a legend: "Identification of sQTLs associated with OSCC risk in normal oral epithelial tissue"; subpanel 1a is captioned "Differentially expressed transcripts between 81 pairs of OSCC and adjacent normal oral epithelial tissues."

Notes

The numerical discrepancy (67 vs. 81) may, in principle, reflect two distinct cohorts (DEG vs. sQTL), but the legend placement and the abstract summary do not make this distinction explicit, creating a plausible source of reader confusion.
Vertical white-line artifacts in blots can arise from legitimate splicing and from other imaging steps; raw full-membrane scans are required to determine origin.
The report's AI-assisted detection flags visual inconsistencies only; formal adjudication of misconduct requires an institutional or editorial investigation.
All numeric evidence and the DOI above are reproduced exactly from the source report.