Dr. Bik is a microbiologist who has worked at Stanford University and the Netherlands National Institute of Health.
One evening in January 2014, I was sitting at my computer at home browsing research papers. As a microbiologist, this was not unusual, although I certainly did not expect to find what I did that night.
These particular articles were descriptions of medical research, and many of them contained pictures of biological samples such as tissues. One photo caught my attention. Was there something familiar about it? Curious, I quickly looked through other articles by the same authors, comparing their images with each other.
It was. An excerpt of the same photo was used in two different articles to present the results of three completely different experiments.
Moreover, the authors appeared to be deliberately blurring their tracks. Although the photos were from the same sample, one appeared to be turned backwards, while the other appeared to be differently stretched and cropped.
Two papers, three experiments, one image
These figures show western blots, which are used to detect the presence of a specific protein in tissues or bodily fluids.
Sources: “REDOX regulation of IL-13 signaling in intestinal epithelial cells: the use of alternative pathways mediates distinct patterns of gene expression”, Debasmita Mandal, Pingfu Fu and Alan D. See the article : Central College adds data science options | Local News | oskaloosa.com. Levine (first work), “Elevated IL-13Rα2 in intestinal epithelium of cells with ulcerative colitis or colorectal cancer initiate the MAPK pathway ”by Debasmit Mandal and Alan D. Levine (2nd work).
Even though that was eight years ago, I remember perfectly well how angry it made me. It was a scam, clean and simple. By editing an image to obtain the desired result, a scientist can create evidence for a favorite hypothesis or create a signal from noise. Scientists need to rely on and build on their work. Cheating is an offense against all that science should be. If scientific articles contain errors or – worse – false data and fabricated images, other researchers will likely waste their time and spend money chasing theories based on fabricated results.
But were these duplicate images just an isolated incident? Having no idea how big it would be, I started looking for suspicious figures in biomedical journals.
From childhood, I was “blessed” with what I was told was a better than average ability to spot repeating patterns. It’s a doubtful blessing when you focus more on the floor tiles than on the person you should talk to. However, this ability, combined with my – what some might call obsessive – personality has helped me look for duplicate scientific images by eye.
During the day I went to work in the lab at Stanford University, but soon spent every evening and most weekends looking for suspicious images. In 2016, I published an analysis of 20,621 peer-reviewed articles, uncovering problematic images in at least one in 25. Half of them appeared to be deliberately manipulated – rotated, inverted, stretched, or otherwise altered. Feeling anxious about how much bad science could be in journals, I quit my full-time job in 2019 to be able to dedicate myself to finding and reporting more cases of science fraud.
Using pattern-matching eyes and loads of caffeine, I analyzed over 100,000 articles since 2014 and found visible duplication of the image in 4,800 and similar evidence of error, fraud, or other ethical issues in an additional 1,700. I reported 2,500 of them to their journal editors and – after finding out it seems that journals often fail to respond to such cases – I have published many of these articles, along with 3,500 others, on PubPeer, a website where the scientific literature is discussed publicly. .
While some of these studies may be relatively insignificant, not all of them. Earlier this year, Science magazine asked me to comment on seemingly manipulated photos featured in influential Alzheimer’s research at the University of Minnesota. The article claimed to offer unique evidence of the underlying cause of Alzheimer’s disease.
Matthew Schrag, a neuroscientist and physician at Vanderbilt University, has already found dozens of suspicious images in articles by one of the researchers, Sylvain Lesné. While checking his findings, I agreed and found even more. (A representative from the University of Minnesota said the university was analyzing questions about his work.)
Other researchers have not been able to recreate the famous University of Minnesota research. Now that the images in these articles show signs of deliberate manipulation, it raises questions across a whole line of research, with potentially millions of dollars in wasted grant money and years of false hope for patients. However, not all can be completely lost; Pharmaceutical companies Biogen and Eisai recently said that they are developing an anti-amyloid drug for Alzheimer’s disease.
In 2018, Harvard Medical School and Brigham and Women’s Hospital in Boston charged former employee Dr. Piero Anversa and his laboratory for falsifying or fabricating data and images in 31 scientific papers over almost two decades. Dr. Anversa’s lab pioneered the theory that bone marrow stem cells could regenerate the human heart by injecting it into it. According to a Reuters analysis, the National Institutes of Health has spent at least $ 588 million continuing this line of research. Other scientists have never been able to replicate his amazing results. Dr. Anversa blamed a colleague and said he was unaware there was fraud in his lab.
Just last month, geneticist Nobel laureate Gregg Semenza had to withdraw four of his articles after revealing that they contained images that appeared to be manipulated or duplicated. The prestigious Proceedings of the National Academy of Sciences published the withdrawn articles. The discovery was made by the pseudonym Clare Francis, the “science detective.”
Elisabeth Bik in her home office.
Amy Osborne / Agence France-Presse – Getty Images
Most of my fellow detectives remain anonymous, operating under pseudonyms such as Smut Clyde and Cheshire. Criticism of the work of other scientists is often not well received, and concerns about the negative consequences of a career can prevent scientists from speaking out. Image issues that I reported under my full name resulted in hate messages, angry videos on social networks and two threats of lawsuits.
The Times tried to contact the lead scientists of the retracted articles with the images reprinted in this essay. Only one replied; others did not reply or declined to comment. The author who replied, Thomas J. Webster, said posting the photos was an honest mistake.
Of course, the images themselves do not reveal directly how they were created or what authors were involved in their creation. While some duplicate images appear to be the result of deliberate editing, it is possible that others were the result of sloppy lab work, accidental mislabelings, or misunderstandings between colleagues.
There are several things that can trick scientists into cheating. First, most scientists feel pressure to publish. Publications are essential for a researcher’s career and are critical to an academic position. Employers can request a certain number of published articles over time, pay bonuses, or promote employees after publication. Overall, studies that show successful results are more likely to be published than those that fail to confirm the hypothesis. So when a scientist’s research is negative, deception can be tempting. Or perhaps a scientist has received praise and attention in the past for a significant discovery, but has entered the idle of research. In these cases, they may be tempted to “adjust” their results to make them appear more convincing. Some laboratories are run by over-demanding – and perhaps even intimidating – professors. As a result, to receive a letter of recommendation that will enable them to flee to a new position, young researchers may be desperate to please.
Before scientific articles are published, they are peer-reviewed, a process by which two or three independent scientists evaluate the article for scientific rigor and correct analysis. But peer-review is free and underrated, and the system is built on a trusted, non-contradictory relationship. Peer review is not configured to detect fraud.
Often times, problems with data – tables, statistical tests, graphs, and photos – are only noticed after publication, when the study is read by a much wider audience. Minor errors can be corrected with a correction. But the article should be withdrawn if critics can demonstrate scientific misconduct, such as taking pictures or falsifying data. After retirement, it will still be available for reading or download, but will be marked as untrustworthy.
Unfortunately, many scientific journals and academic institutions are slow to respond to evidence of image manipulation – if they take action at all. So far my work has resulted in 956 corrections and 923 references, but most of the articles I have submitted to the journals remain unaddressed.
Scientific publishers care enormously about their reputation, and research institutions may be ashamed to admit that there has been improper conduct within their walls. When confronted with duplicate images, they often conclude that there are “mistakes”. Institutions may dismiss some of the junior researchers, but lab managers usually remain in place.
Things can get even worse. Artificial intelligence can help detect duplicate data in research, but it can also be used to generate false data. Nowadays, it is easy to produce fabricated photos or videos of events that never happened, and images generated by artificial intelligence may already have started poisoning the scientific literature. As technology advances, it will be much harder to tell a counterfeit from reality.
Science must take research fraud seriously. Magazines should withdraw articles that contain altered photos or manipulated data much faster – and shouldn’t publish them in the first place. Scientists who find errors in published results should not be threatened with lawsuits to silence criticism.
Here is a list of things that I think need to change:
Journals must perform better quality control. Publishers should hire image analysts and statistical experts to review approved articles prior to publication.
Magazines need to run much faster – within six months, for example – when evidence of image manipulation becomes apparent.
