Integrity Review Report: Microbiota-mediated nitrogen fixation and microhabitat homeostasis in aerial root-mucilage (Microbiome, 2023)

Academic fraud report · Geng Detector

Summary

This automated integrity review assesses Pang et al. (2023), DOI 10.1186/s40168-023-01525-x, published in Microbiome. Overall verdict: highly suspicious. Four findings were confirmed on review: (1) contradictory sample sizes within a single Figure 3 caption (six independent experiments vs n=3); (2) conflicting 15N2 gas purity specifications (≥98% in Results vs 99.9% in Methods); (3) statistically anomalous last-digit distributions across the full text (chi-squared=274.6, p<0.0001; odd/even ratio 12:69, p<0.0001; adjacent-digit correlation 11.3 sigma), suggesting possible numerical manipulation; and (4) ambiguous sample sizes and statistical reporting in the 15N natural abundance experiment (12/8 samples vs n=6). Three additional findings were retained as unresolved concerns: identical error bars across groups (Fig 2b, 3b–e), copy-move forensics positives in 7 embedded images, and suspiciously clean petri-dish photos (Fig 5). These remain unverified pending original data. Confidence in confirmed text-based inconsistencies is high (verbatim verified); image-based findings carry moderate uncertainty due to possible benign explanations (multi-panel chart layouts). Recommend editorial investigation and author data disclosure.

Verdict

Highly suspicious (orange/red). Four independently confirmed anomalies after review (sample-size contradictions, reagent purity contradiction, last-digit statistical anomalies, ambiguous natural-abundance reporting), plus three downgraded concerns retained (image forensics positives, identical error bars, overly clean petri-dish photos). The aggregation of multiple independent anomalies warrants formal editorial and institutional investigation. This is not a definitive finding of misconduct.

Key findings

Finding 4 (confirmed): Figure 3 caption states "six independent experiments" yet also reports "(P < 0.05, two-way ANOVA, n = 3)" — a direct internal contradiction in the paper's core quantitative experiment. Text also lists 5%, 10%, and 20% ¹⁵N₂ dilutions while the caption describes only 20%.
Finding 5 (confirmed): ¹⁵N₂ gas purity stated as "≥ 98%" in Results vs "99.9% purity" (Wuhan Isotope Technology) in Methods — a key parameter conflict for isotope quantification.
Finding 6 (confirmed, last-digit portion): Last-digit distribution severely non-uniform (χ²=274.6, p<0.0001); odd/even ratio 12/69 (p<0.0001); adjacent last-digit correlation 11.3σ. Benford tests were rejected due to unsuitable data types. Verified regular values include "59.09%, 10.43%, and 5.36%", "0.14%, 0.22%, and 0.64%", "37.04 to 54.85%".
Finding 7 (confirmed): Methods report "nearly 12" mucilage samples and 8 reference plants (2020–2021), while Fig 3 caption uses "(P < 0.01, T test, n = 6)"; two-year data reported only as merged range "37.04 to 54.85%".
Finding 1 (unverified): Error bars in Fig 2b and Fig 3b–e appear visually identical across groups; requires underlying data.
Finding 2 (unverified): Copy-move positives in 7 images (e.g., img-006 offset (0,32), 93 block pairs; img-007 offset (0,32), 45 pairs; img-011 offset (88,0), 11 pairs; img-010 offset (32,0), 18 pairs). May partly reflect repeated axes in multi-panel figures; img-011 shows high noise variance (CV=1.84).
Finding 3 (unverified): Fig 5a–b petri-dish photos appear unusually clean for 5-day open-air exposure; possible over-processing — requires originals.

Evidence highlights

DOI: https://doi.org/10.1186/s40168-023-01525-x (Microbiome 2023; 11:85; received 2022-11-30, accepted 2023-03-20).
Internal contradiction: "six independent experiments" vs "n = 3" within the same Fig 3 caption (verbatim verified).
Purity conflict: "purity specification ≥ 98%" (Results) vs "99.9% purity" (Methods); the Results phrase was missed by automated matching due to line-break hyphenation but confirmed manually.
Last-digit statistics: χ²=274.6, p<0.0001; odd/even 12/69; adjacent-digit deviation 11.3σ.
Bayesian aggregation (programmatic, prior fixed at 5%): posterior ≥99.9%, BF = 6.88e+6; assumes conditional independence and may overstate correlated evidence; 11 leads excluded on review as benign/unfounded. Not a misconduct determination.

Notes

Recommended actions: request raw data (¹⁵N IRMS reports, natural-abundance records, original Fig 5 photos, Fig 2/3 underlying data), post on PubPeer, notify Microbiome/BMC editorial office for image forensics on Fig 3 caption, Fig 5, and the img-011 page, and inform institutional committees (Xishuangbanna Tropical Botanical Garden, CAS; Zhejiang University).
Image-based and visual findings remain unconfirmed pending original high-resolution files; benign explanations have not been ruled out.
AI-assisted report for academic discussion; final determination rests with formal institutional investigations.