Geng report: Long non-coding RNAs are involved in alternative splicing and promote cancer progression (British Journal of Cancer, 2022)

Academic fraud report · Geng Detector

Summary

Verdict: Questionable. A focused review by Ouyang, Zhong, Zhang et al. published in British Journal of Cancer (DOI 10.1038/s41416-021-01600-w) was examined for internal consistency and image integrity. The report identifies one substantive internal contradiction and several machine-forensic signals that could not be confirmed. The headline issue is Finding 1: PXN-AS1-S is described in the main text (Fig. 2a narrative) as the exon-4-lacking isoform that inhibits liver cancer, with MBNL3 as the factor inducing exon 4 inclusion to produce the long isoform PXN-AS1-L. Table 1 instead assigns PXN-AS1-S to SRSF1-driven exon 4 skipping and labels it pro-tumorigenic, directly inverting the isoform's role and splicing factor within the same paper. Findings 3 and 4 (Fridrich copy-move matches and noise-variance CV values in Fig. 1–4) are downgraded because those figures are vector-style mechanistic schematics lacking raw experimental content, so the algorithm cannot distinguish benign template repetition from manipulation. The submission timeline (Received 6 Dec 2020; Accepted 11 Oct 2021) is consistent. Limitations: only textual and machine-forensic signals are reported; no human image re-inspection was performed, and the journal's own editorial investigation is required for a definitive judgment.

Verdict

🟡 Questionable. One reproducible internal contradiction in the text/table is documented; image-forensic signals are present but have plausible benign explanations. No definitive fraud finding.

Key findings

Finding 1 (substantive): Direct contradiction for PXN-AS1-S between Fig. 2a narrative and Table 1 — opposite functional role (tumor-suppressive vs pro-tumorigenic) and different splicing factor (MBNL3 vs SRSF1).
Finding 2 (not substantive): PVT1ΔE4 is consistently described across Table 1 and Fig. 2b; only phrasing granularity differs.
Finding 3 (downgraded): Fridrich block-matching detected regular copy-move signals in Fig. 1–4 (offsets (48,0), (96,0), (72,0), (0,32); 8, 27, 57, and 37 matched pairs respectively). All four panels are vector schematics without raw experimental imagery, so the matches likely reflect reused template elements rather than paste-tampering.
Finding 4 (downgraded): Noise-variance CV values 0.89 / 0.88 / 0.96 / 0.81 (with 6 and 21 anomalous blocks in img-002 and img-003) are not indicative of sensor-noise inconsistency because the figures are vector schematics lacking sensor-captured content.
Finding 5: Timeline and citation dates (Received 6 December 2020; Revised 14 September 2021; Accepted 11 October 2021; Published online 8 November 2021) are internally consistent; no temporal anomalies.

Evidence highlights

DOI: 10.1038/s41416-021-01600-w.
Fig. 2a narrative: "MBNL3 ... induces lncRNA-PXN-AS1 exon 4 inclusion ... The transcript of lncRNA-PXN-AS1 lacking exon 4 ... inhibiting the translation of PXN mRNA and inhibiting liver cancer."
Table 1, PXN-AS1-S row: "splicing factor SRSF1 promotes the exon 4 skipping of PXN-AS1 and generates this isoform ... Inhibits the translation of PXN, thus promoting tumorigenesis."
Table 1, PVT1ΔE4 row vs Fig. 2b: both attribute exon 4 skipping to SRSF1; only presentation granularity differs.
Machine-forensic reports [I1]–[I8]: documented pixel offsets, matched-pair counts, CV values, and anomalous-block counts listed above; all four embedded images are schematic in nature.
No ethics approval number to verify (review declares Not applicable); no reagent or equipment identifiers requiring cross-check.

Notes

The contradiction in Finding 1 is a clearly verifiable text-vs-table mismatch within a single article and is the report's primary actionable concern.
Image-forensic results on schematic figures carry limited discriminatory power and should not, on their own, be treated as evidence of manipulation.
Recommended follow-up: flag the PXN-AS1-S inconsistency on PubPeer and request an Erratum from the authors and the journal; no further image-focused investigation is warranted for this paper.