A March 2026 U.S. Office of Research Integrity (ORI) case summary against a former Icahn School of Medicine at Mount Sinai postdoctoral fellow has, in the months since, produced a second-order effect increasingly common in research-integrity cases: an ORI finding triggering independent image-forensics sleuthing that surfaces an unrelated instance of misconduct in a published, cited paper, which then leads to a formal journal retraction. That is what happened this year with Chen-Yeh “George” Ke, PhD.
The original ORI finding
ORI’s case summary, posted to ori.hhs.gov in March 2026, found that Ke, a former postdoctoral fellow in the Department of Cell, Development and Regenerative Biology at the Icahn School of Medicine at Mount Sinai (ISMMS), engaged in research misconduct under 42 CFR Part 93. The finding covered one draft manuscript (‘Transient activation of ERK promotes cell differentiation through non-canonical Wnt Signaling’) and two NIH Research Performance Progress Reports (RPPRs) submitted to the National Institute of Dental and Craniofacial Research (NIDCR) for grant R01 DE022363.
ORI found by a preponderance of the evidence that Ke intentionally and knowingly falsified and/or fabricated western blot images — including cut-and-pasted bands, duplicated lanes with horizontal flipping, and vertical shifting — along with duplicated and relabeled microscopy images of alkaline-phosphatase-stained cells, and duplicated and flipped calcium-imaging data with altered brightness, used to misrepresent different experimental conditions across figures. The administrative actions, which Ke did not contest within the required 30-day window, impose three years of supervision for any PHS-supported research (committee oversight, quarterly primary-data review, advance manuscript review, and six-month reporting to ORI) and a three-year prohibition from serving in any advisory capacity to the U.S. Public Health Service.
Retraction Watch first reported the finding on March 11, 2026, noting Ke was no longer at Mount Sinai and, per LinkedIn, was working as a manager at a biotechnology company in Taiwan at the time.
How the finding led to sleuthing on unrelated, published work
The ORI case summary itself covered only the unpublished manuscript and the two RPPRs — it made no finding about any of Ke’s published papers. But ORI case summaries, once public, routinely draw the attention of independent research-integrity sleuths who treat a confirmed misconduct finding as a reason to re-examine a respondent’s full publication record, on the reasoning that a researcher found to have manipulated images in one context is worth checking for the same pattern elsewhere.
That is what happened here. After Retraction Watch’s March report, image-integrity researcher Paul S. Brookes reviewed Ke’s ORCID record and ran Ke’s published papers through ImageTwin.ai, an automated image-forensics tool that screens figures against a reference database of scientific images spanning more than 160 million published works to flag duplication and reuse. The screen surfaced a match between a figure in a 2019 Science Signaling paper — first-authored by Ke — and a figure in an earlier, unrelated 2013 Scientific Reports paper, also first-authored by Ke. Specifically, an image the 2019 paper presented as a wild-type mouse sample matched an image used in the earlier work to represent virus-infected mouse tissue — meaning the same underlying image had been presented as two different biological conditions in two separate publications. Brookes reported the duplication to Science Signaling and posted it to PubPeer on March 13, 2026.
The Science Signaling retraction
The journal’s subsequent investigation confirmed the reported duplication and, per Retraction Watch’s follow-up reporting, turned up additional image duplications and discrepancies in the same paper beyond the one Brookes had flagged, including further duplicated western blots and inaccurate figure labeling. The corresponding authors requested retraction after the investigation confirmed the findings, and the 2019 paper — which had accumulated roughly 20 citations — was formally retracted in July 2026.
Retraction Watch’s account of the episode, published July 14, 2026, frames it as a now-familiar pipeline: a federal misconduct finding becomes public, independent sleuths use automated image-forensics tools to check the respondent’s broader publication record for the same failure mode, and a publisher retraction follows months later on a paper the original federal finding never touched.
Why this matters for research-integrity and compliance workflows
The Ke case is a concrete illustration of a pattern research-integrity officers (RIOs), compliance offices, and CASRAI’s own research integrity stakeholders should expect as standard practice rather than a one-off: an institutional or federal misconduct finding scoped narrowly to specific grant-reported data or manuscripts does not reliably contain the reputational or bibliographic fallout to that scope. Once a respondent’s name and institution are public via an ORI case summary, independent tools such as ImageTwin and Proofig, combined with community sleuthing on PubPeer, routinely extend scrutiny to the respondent’s full publication history — including papers with no connection to the original federally funded work, and papers published years before or after the conduct that triggered the original finding.
For institutions, this has practical implications: a narrow finding against a former trainee or staff member can still generate follow-on retraction requests months or years later, on papers the institution may not have separately reviewed, and those retractions can implicate co-authors and institutions who had no role in the original misconduct. For RIOs managing a case, tracking downstream PubPeer/retraction activity connected to a respondent — not just closing the file once ORI’s administrative action is final — is increasingly part of a complete institutional response.
Frequently asked questions
What is the U.S. Office of Research Integrity’s role in a case like this?
ORI oversees institutional investigations into allegations of research misconduct (fabrication, falsification, or plagiarism) in PHS-funded research, reviews the institution’s findings, and can issue its own administrative actions — ranging from supervision requirements and PHS-advisory prohibitions, as in this case, to full debarment in more severe cases. ORI’s own investigation and finding are scoped to the specific research it reviewed; they do not constitute a review of a respondent’s entire publication record.
Did ORI’s finding against Ke cover the 2019 Science Signaling paper?
No. ORI’s case summary addressed only a draft manuscript and two NIH RPPRs tied to a specific NIDCR grant. The Science Signaling retraction arose from separate image-forensics sleuthing conducted independently, after the ORI finding became public, and covers unrelated, already-published work.
What tools are typically used for this kind of post-finding sleuthing?
ImageTwin.ai and Proofig are the two most commonly cited automated image-forensics tools in current research-integrity sleuthing and publisher screening workflows; both compare submitted or published figures against large reference databases to flag duplicated or manipulated images. Findings are frequently cross-posted to PubPeer, the post-publication peer review platform, where they become publicly discoverable and citable by journals conducting their own investigations.
Related CASRAI coverage
See CASRAI’s coverage of the same batch of 2026 ORI case summaries, including the Fernandez and Andrade cases, in ORI’s 2026 Case Summaries: A 15-Year Debarment and What the Fernandez Case Shows About Enforcement, and CASRAI’s dictionary entries on Research Integrity Officer (RIO) and retraction for background on the underlying processes and terminology.
