Suspicious statistical, methodological, and reporting anomalies in 'A battery-free nanofluidic intracellular delivery patch for internal organs' (Nature, 2025)

Academic fraud report · Geng Detector

Summary

This report evaluates a 2025 Nature paper describing a nanofluidic intracellular delivery device for internal organs, co-authored by Dedong Yin, Pan Wang, Yongcun Hao and colleagues. The overall verdict is highly suspicious (🟠), based on multiple but non-visual findings due to PDF text extraction limitations. Key issues: (1) Implausibly extreme P-values in small-sample in vivo experiments, including P = 1.10 × 10^-14 and P < 1 × 10^-15 for serum ALT/AST in mice (n=10, Fig. 3c/3d) and P = 3.22 × 10^-10 for tumour weight with n=5 (Fig. 5e), requiring near-zero variance inconsistent with biological variability. (2) A likely misidentified reagent catalogue number: Sigma F8318 is reported for penicillin-streptomycin, but this number is widely associated with foetal bovine serum rather than pen-strep solutions (commonly P0781 or P4333). (3) An arithmetic inconsistency in single-cell sequencing knockout distributions (3% + 21% + 68% = 92%, with 8% unaccounted cells silently dropped relative to the in vitro profile). (4) An unusually long 2-year-5-month review cycle. Image-level duplication analysis was not possible from text only. Confidence is moderate-to-high for the statistical and percentage anomalies, lower for the reagent error (could be a typo).

Verdict

Highly suspicious (🟠). Multiple convergent concerns across statistics, reagents, and data reporting were identified from text extraction. Image-level forgery checks were not feasible. No definitive proof of misconduct is established; the issues warrant investigation by the journal and the authors' institution.

Key findings

Implausibly extreme P-values across multiple figures in small in vivo cohorts (see Evidence highlights).
Suspected misidentification of a basic reagent catalogue number (penicillin–streptomycin listed as Sigma, F8318, a number commonly associated with foetal bovine serum).
Internal inconsistency in single-cell knockout percentages: reported in vivo distribution sums to 92%, with 8% of cells unaccounted and not discussed.
Extended review timeline of approximately 2 years and 5 months (Received 11 October 2022; Accepted 26 March 2025), unusual for a Nature main research article.
Image duplication/figure manipulation analysis could not be performed because the supplied text lacked pixel-level image data.

Evidence highlights

Figure 2b, 2g, 3c, 3d, 3f, 5d, 5e (Statistical anomaly):
Fig. 3c/3d: serum ALT/AST in mice reported at P = 1.10 × 10^-14 and P < 1 × 10^-15 with n = 10.
Fig. 5e: tumour weight P = 3.22 × 10^-10 with n = 5.
Such extreme significance at these sample sizes would require error bars effectively at zero, which is biologically implausible for in vivo liver-injury and tumour-growth assays and is consistent with data cherry-picking, over-processing, or back-calculation.
Methods / Cell culture (Reagent anomaly): 100 units ml–1 of penicillin–streptomycin (Sigma, F8318). Industry knowledge associates F8318 with foetal bovine serum rather than pen-strep; common Sigma pen-strep catalogue numbers are P0781 and P4333.
Figure 4a and accompanying text (Arithmetic inconsistency): text states 3% Brca1 single-gene KO, 21% Trp53 single-gene KO and 68% Brca1 and Trp53 double-gene KO, totalling 92%. The in vitro comparator shown in the figure lists four categories (8%, 3%, 21%, 68%) summing to 100%; the 8% Unchanged category is omitted from the in vivo description without justification.
Article metadata (Timeline anomaly): Received 2022-10-11; Accepted 2025-03-26; published online 2025-04-30. DOI: 10.1038/s41586-025-08943-x. Review-to-acceptance duration ≈ 29 months.
Image analysis (Inconclusive): No pixel-level evidence was available from the provided PDF text; therefore no conclusions can be drawn regarding Western blot splicing, clone regions, or background noise consistency.

Notes

The reagent catalogue mismatch is a single point and could in principle reflect an honest typo; however, combined with the other anomalies it contributes to an overall pattern of insufficient methodological rigour.
The timeline anomaly may be explained by extensive long-term in vivo work (the paper describes experiments spanning ~60 weeks), but the possibility of retrospective data modification cannot be ruled out without raw data review.
Authors should be requested to provide raw individual-level data and error calculations for Figures 3 and 5, the original single-cell sequencing counts underlying Figure 4a, and the original full-resolution figures/uncropped blots.
Institutional animal ethics oversight records for a 60-week high-mortality tumour experiment are also worth verifying.
This report is AI-assisted and based solely on text extracted from the supplied PDF; image-based checks require the original file layers.