Integrity Concerns in 'Release of a ubiquitin brake activates OsCERK1-triggered immunity in rice' (Nature, DOI 10.1038/s41586-024-07418-9)

Academic fraud report · Geng Detector

Summary

This report assesses a 2024 Nature paper by Wang et al. on OsCIE1-mediated ubiquitination of OsCERK1 in rice. Overall verdict: highly suspicious (🟠), pending verification against original uncropped data. The strongest findings are (1) an apparent vertical seam in the Figure 2a anti-Flag Western blot, with a background intensity discontinuity between lanes 5 and 6, corroborated by automated noise-variance forensics (CV=1.774); (2) suspected reuse of lesion leaf images between two distinct complementation lines (Oscie1-1COM(4D) vs Oscie1-1COM(S237A)) in Figure 3c, supported in direction by copy-move detections; and (3) batch copy-move and noise-uniformity signals across 10+ images. A statistical digit test was flagged but its premises are invalid (mixed non-measurement values). Authors state blots were run on separate membranes, but this does not explain within-membrane discontinuities. Original full membrane scans, raw leaf photos with EXIF, and Source Data are required before any conclusion of misconduct; the automated Bayesian posterior (≥99.9%) should be treated with caution given conditional-independence assumptions.

Verdict

Highly suspicious (🟠) pending verification with original, uncropped source images. The report identifies suspected lane splicing in Figure 2a and possible image reuse in Figure 3c, supported in direction by programmatic forensics. However, original gel scans and raw photographs (Supplementary Information) were not available for direct inspection, so visual conclusions require confirmation with original files. No timeline or reagent provenance red flags were found.

Key findings

Figure 2a (Western blot, anti-Flag panel, p. 1160): Apparent vertical seam between lanes 5 ('+') and 6 ('m', OsCIE1(C234A/W261A) negative control) — left side darker/uniform, right side markedly brighter. Independent forensic analysis of img-030.jpg flagged noise-variance heterogeneity (CV=1.774) → suspected splicing (I11). The Methods statement that immunoblots were run on separate membranes in parallel does not explain within-membrane background discontinuities. Severity: 🟠, confirmed on review.
Figure 3c (M. oryzae leaf phenotype photos, p. 1161): First leaves of 'Oscie1-1COM(4D)' and 'Oscie1-1COM(S237A)' — genetically distinct independent lines — show highly similar lesion shapes, vein patterns, and edge defects. Copy-move forensics found same-offset matches on related page images (26 pairs, 18 pairs; I27, I39). Visual confidence is moderate and copy-move regions are not yet pixel-localized to the leaves; retained as a suspicion requiring original photos with EXIF. Severity: 🟠, unresolved concern.
Figure 2d (³²P autoradiography, p. 1160): Mild background discontinuity between lanes 3 and 4; alone insufficient for any conclusion. Severity: 🟡, unresolved.
Statistical digit tests: Last-digit χ²=1277.7, odd/even ratio 26/206, adjacent-digit correlation 30.7σ; Benford deviations reported. However, extracted values include axis ticks, Y2H labels, and panel letters, so test premises are not met — not independent evidence of fabrication. Re-analysis should use Source Data measurement values. Severity: 🟡, insufficient basis.
Batch forensic signals: Copy-move matches and noise heterogeneity (CV=1.65–2.08) on 10+ images (🚨-level signals at offsets (0,48), (0,32), (136,0), etc.). Known benign causes exist for multi-panel layouts, but the consistency with the visual findings in Figs. 2a and 3c warrants original-file verification.
Timeline check: Received 2022-04-11, accepted 2024-04-12, published 2024-05-15; PDB (7XED, 8HQB, 8HPB) and BMRB (36528, 36527) entries present. No timeline red flags; the disclosed single MS analysis is a limitation, not evidence of misconduct.

Evidence highlights

Forensic: noise CV=1.774 at Fig. 2a page (img-030.jpg, I11); copy-move pairs at offsets (0,32), (0,40), (0,48), (136,0) across img-001, img-002, img-030, img-033, img-047, img-069, img-1052, img-1057, img-1059 (10 🚨-level, 20+ ⚠️-level signals).
Top Bayesian contributions: I2 (20 pairs, offset (0,48), logLR=1.46, 9.9%), I39 (18 pairs, (0,40), 9%), I29 (9 pairs, (0,32), 8.8%), I3 (14 pairs, (0,32), 8.2%), I36 (10 pairs, (136,0), 7.3%).
Reported composite posterior ≥99.9% (BF = 1.1e+8) is auto-generated, assumes conditional independence of evidence, and must not be read as a misconduct determination; 33 clues were excluded on review.

Notes

Original Supplementary Information gel blots were not obtained; authors state "Original gel blots are shown in Supplementary Information Fig. 1."
Recommended actions: request original uncropped membranes (Figs. 2a/2d), raw Fig. 3c leaf photos with EXIF, and Source Data for digit/GRIM re-testing; post comparison images on PubPeer; escalate to Nature editors and the authors' institution (CAS Center for Excellence in Molecular Plant Sciences) as warranted.
AI-assisted report for academic discussion only; final misconduct determinations require institutional investigation. False positives and negatives are possible.