Geng Academic Fraud Scan Report — Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage (Komor et al., Nature 2016)

Academic fraud report · Geng Detector

Summary

Verdict: ✅ Cleared. The review of Komor et al.'s foundational base-editor (BE3) paper finds no substantive evidence of academic fraud. Image-level forensic analysis (duplication, splicing) was not possible because high-resolution original figures were not supplied; this is a methodological limit, not a positive signal of misconduct. Text-based scrutiny across six dimensions revealed no anomalies. High-throughput sequencing data in Figure 4 (e.g., APOE4 BE3-treated reads at target C5: A 0.1, C 0.0, G 0.9, T 74.9) show natural low-frequency background noise consistent with genuine sequencing artefacts rather than fabricated values. Statistical reporting is transparent: the authors openly state no a priori sample-size determination and no blinding, and present data as mean ± s.d. with n=2 or n=3, including honest single-experiment designations. The submission timeline (received 26 Feb 2016; accepted 30 Mar; online 20 Apr; print 19 May) is consistent with a landmark, rapidly reviewed Nature paper. Methods, reagents (VeraSeq Ultra, Q5 polymerase, Lipofectamine 2000), cell lines (HEK293T, U2OS, HCC1954), Illumina MiSeq sequencing, SRA accession SRP072434, and Addgene plasmids 73018–73021 are all plausible and internally consistent. Confidence: high for non-image checks; low for image-related claims.

Verdict

✅ Cleared — no credible evidence of academic fraud detected within the limits of a text-based review.

Key findings

Image-level analysis not feasible: Original high-resolution images were not available, so pixel-level checks for duplication or splicing could not be performed. This is a constraint, not an indication of misconduct.
Sequencing data appear genuine: Reported HTS percentages in Figure 4 (APOE4 and TP53) display natural low-frequency noise (e.g., values such as 0.1, 0.9, 0.7) consistent with real sequencing background, rather than the suspiciously clean distributions typical of fabricated data.
Transparent statistical reporting: Methods explicitly disclose the absence of a priori sample-size calculation and no blinding. Biological replicates (typically n=2 or n=3) are reported in figure legends, and single-experiment results are honestly labeled as such (e.g., Extended Data Fig. 4b).
Plausible and consistent timeline: Received 26 Feb 2016; accepted 30 Mar 2016; published online 20 Apr 2016; in print 19 May 2016 — consistent with rapid review of a high-impact, well-prepared manuscript.
Methodological detail is verifiable: Reagents (VeraSeq Ultra, Q5 polymerase, Lipofectamine 2000), cell lines (HEK293T, U2OS, HCC1954), Illumina MiSeq platform, SRA accession SRP072434, and Addgene plasmids 73018–73021 are all consistent with 2016-era practice and provide concrete traceability.

Evidence highlights

Specific sequencing readout preserved: APOE4 BE3-treated target C5 — A 0.1, C 0.0, G 0.9, T 74.9 (Figure 4 text panel).
Authors direct readers to uncropped gel source data in Supplementary Figure 1, demonstrating open data-hygiene practice.
DOI: 10.1038/nature17946 (Nature, Vol. 533, 2016).
Cited references include contemporaneous 2016 literature (refs 11, 28), with no temporal inconsistencies.

Notes

Limitations: Without access to the original figures (raw Western blots, gels, microscopy), any conclusion about image manipulation is necessarily incomplete. Independent verification by image-forensics tools or institutional review is recommended for definitive adjudication.
The flagged openness about sample size and blinding is best practice in current reporting norms, even where it deviates from later ARRIVE-style recommendations.
This report is AI-assisted and intended for academic discussion only; it does not constitute a formal institutional finding.