Just as a Harvard lab brought in tens of millions of dollars in private equity funding to pursue new treatments for obesity, past research from its lead investigator has come under fresh scrutiny.
Last month, the lab of Gökhan Hotamışlıgil, a professor of genetics and metabolism at Harvard’s T.H. Chan School of Public Health, secured a $39 million dollar investment from İş Private Equity, an Istanbul-based firm. The partnership centers on FABP4, a protein associated with obesity and other metabolic conditions.
But over the past decades, two of Hotamışlıgil’s papers have been corrected for image duplications, and since the announcement, renewed scrutiny of Hotamışlıgil’s work appeared on PubPeer, including for issues with statistical analyses.
“Of course when we become aware of an irregularity or mistake, or someone makes an allegation, we take these issues very seriously and do all we can to address the question,” Hotamışlıgil told us in an email.
The most recent comments are from Reese Richardson, a computational biologist who looks into scientific reproducibility. Richardson said he became curious about Hotamışlıgil’s work after seeing news about the funding deal. He soon found existing concerns flagged on PubPeer.
“I figured that it would be worthwhile to check his papers out,” Richardson told us. Initially, two articles piqued his interest, although Hotamışlıgil is not the senior author on either.
In one PubPeer post, Richardson raised concerns about a 2017 paper in Nature Medicine in which he discovered mislabelling of data, claims of statistical significance the raw data didn’t seem to support, and some missing source data. The author contribution statement indicates Hotamışlıgil oversaw some of the experiments.
Richardson also looked into a 2019 article in Cell Metabolism that contained “either wrongly reported or apparently manipulated” statistics, he told us. Nearly every result “features inexplicable methodological inconsistencies or reporting errors,” he wrote on PubPeer. “These result in a massive exaggeration of the effects observed.”
According to Richardson’s analysis, some statements made in the paper were true only when samples were arbitrarily excluded, but the methods didn’t describe any procedure for the exclusion of samples. According to the author contribution statement, Hotamışlıgil helped with the discussion and interpretation of results.
James Heathers, a scientific sleuth and director of the Medical Evidence Project — an initiative of the Center for Scientific Integrity, the parent nonprofit of Retraction Watch — said the extent of the problem is mostly unknowable. “How can you adjust for the effects of data that isn’t presented?” If outliers are left in the dataset, but redacted from the analysis, “the redaction should be explained in 100% of cases,” he said.
Hotamışlıgil said the work for those two papers was led by Yu-Hua Tseng, another researcher at Harvard, and his group’s involvement was limited. “We have provided some help with their experiments, my fellow Alex Bartelt provided some training and physiology and biochemistry experiments,” he told us. “We were not directly involved with the rest of the paper.”
Apart from Richardson’s recent comments, anonymous commenters on PubPeer have pointed out irregularities in Hotamışlıgil’s published work over the years. A 2008 PLOS One paper, for example, was corrected after a commenter found a duplicated image. A similar issue was flagged in a 2020 paper in Cell Metabolism, which included two identical images in one figure. “We deeply regret this unfortunate error,” one of the authors wrote on PubPeer. That paper was also corrected. Another flagged paper appears to contain similar-looking blots. Hotamışlıgil is listed as the senior and corresponding author on each of these papers.
For a 2019 article in Science Translational Medicine related to FABP4 — the protein at the center of the new funding deal — an anonymous commenter questioned the statistical significance of the findings. After analysing the raw data, the commenter found no statistical significance for some findings reported in the paper, despite them having been reported as such. Hotamışlıgil was the senior and corresponding author for this paper. Heathers said these statistical concerns comprised “a concerning pattern of oversights” and would “necessitate exactly the kind of audit that universities tend not to do.”
Reflecting on the PubPeer posts appearing under his name, Hotamışlıgil said, “In many cases, there is no basis, for example there are several posts where there is no comment, there is even one post about a review that I have written but nothing can be found at the post, and some state opinions or perceptions.”
Hotamışlıgil also co-authored a 2002 Nature paper with Michael Karin, a prominent cancer researcher and the subject of previous scrutiny for image problems, including in our own reporting. The paper was corrected in 2023 after PubPeer commenters identified duplicated blots. In the correction, the editors wrote: “as the raw data for the blots used in the manuscript are no longer available, unfortunately we cannot ascertain the issue with the figure.” Hotamışlıgil was listed as the senior and corresponding author on this paper.
He has also co-authored papers with Umut Ozcan, who in 2015 was accused by postdocs of fabricating data and creating a hostile work environment. The case was dismissed from court in 2018. Hotamışlıgil was Ozcan’s graduate advisor, and the two articles where Hotamışlıgil is listed as the senior and corresponding author, and Ozcan as the first, appear to contain duplicate lanes in the blot data or have a mismatched number of lanes. Mike Rossner, an image manipulation consultant, said these PubPeer allegations also had merit, although the mismatched lanes could be a clerical error, which is “unusual, but not unheard of,” he added.
Harvard has described the Turkish firm’s partnership with Hotamışlıgil’s lab as a “potential model” for revenue for the T.H. Chan School. According to a press release about the funding, the school has seen almost $200 million in federal funding dry up in recent months, and the Trump administration terminated nearly every direct federal grant for research and training. The deal was in progress before the administration came into office, and would have taken place despite the recent decline in federal funding, a Harvard spokesperson told us.
Iş Private Equity is currently the only private equity funder at the T.H. Chan School, although the school is “accelerating” efforts to engage with the private sector, the spokesperson said.
Such arrangements between universities and private equity are “not common,” Robert Field, a professor of law and public health at Drexel University in Philadelphia, told us. In recent years, private equity has been moving into health care — buying up hospitals, nursing homes and physician practices — but investing into universities is unusual. Private equity is “by nature, profit oriented, and usually short-term profit oriented,” Field said, and are most likely to fund projects that are close to fruition, he said.
The research has the same oversight, regardless of the funding source, said the Harvard spokesperson, and they will not accept collaboration that “imposes any restriction on our faculty’s freedom to publish and speak about their research.”
Richardson told us he shifted his focus to Tseng, who was the senior author on the two recently flagged papers. Richardson came across a different pair of 20-year-old papers she was the lead author on, both of which seem to contain similarities between parts of Western blots, or inconsistent features in the blot lanes.
Rossner told us the allegations about image duplications had merit and the data raise enough concerns for a journal editor to want to verify the published results. Some of the PubPeer comments note mismatched lanes in the Western blots; this could be due to clerical error, Rossner said, and one image was too low resolution to tell whether it could be a duplicate.
Richardson has also since raised concerns about several other papers by Tseng, citing statistical irregularities.
Tseng told us she is aware of the comments and has “already discussed them with the leading authors.” Since speaking with Retraction Watch, she has responded on the platform, writing, “Thank you for your comments. We are currently reviewing the original data and conducting additional analysis. We will respond to the comments once the investigation is complete.”
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on X or Bluesky, like us on Facebook, follow us on LinkedIn, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
Re: “Reflecting on the PubPeer posts appearing under his name, Hotamışlıgil said, “In many cases, there is no basis, for example there are several posts where there is no comment, there is even one post about a review that I have written but nothing can be found at the post, and some state opinions or perceptions.” ”
Dr. Hotamışlıgil is correct that not all PubPeer threads contain a comment (they are essentially “blank”) and many contain comments that are not concerns. This is a reason not to merely count the number of PubPeer threads and say, “this author has 22 comments on PubPeer!” The number of comments does not necessarily correspond with the amount of concerns related to papers on which that author is listed.
But most people reading comments on PubPeer are already familiar with those matters and this isn’t something that Dr. Hotamışlıgil needs to worry about.
It would be desirable if he would review the threads where concerns *are* shared and, regardless of whether he is the corresponding author or not, help the author group address the concerns. All authors on a published paper have represented that they support the entire manuscript, not merely the segments for which they were primarily responsible. One would hope the head of a lab at Harvard would be well aware of that.
One of the main issues identified in the pubpeer comments is failure to properly adjust for multiple comparisons when analyzing large datasets. This is true when hypothesis testing but not if the data are treated as exploratory. In some of the cases identified “hits” generated by analysis of these large datasets without aggressive multiple comparisons adjustments are then validated using what I read to be replicate experiments and orthogonal analytical methods. There are of course additional issues raised in the pupeer comments that go beyond the multiple comparisons testing concerns.
Harvard should be encouraged to scan foreign countries and stop taking the public dole from the U.S. federal government, e.g. N.I.H.
That was a spelling mistake on my part. An unparliamentary verb becomes an verb acceptable in polite society.
I think some Universities are going to be in a world of hurt due to taking PE money. There are often surprisingly few consequences for researchers putting out fraudulent/compromised work via the federal and state governments, but with PE comes someone wanting to recoup their investment if they’ve been mislead, which means lawsuits. With lawsuits comes discovery, and large amounts of potential negligence.
Some of the coauthors of the papers under scrutiny worked for a biotech/pharma company called BERG that was subsequently acquired by a different company.
https://bpgbio.com/