Spot-the-duplicate: the rise of fake images in science
Katie Fegan discusses the impact of image manipulation in academic papers, and the whistleblowers fighting for scientific integrity.
Move over, Sherlock: the science detectives are in town.
Bad science threatens the credibility of academic research. Whether you are an astrophysicist or a microbiologist, chances are you have stumbled across a research paper containing fake images – probably without even realising it. At best, the figures might have been incorrectly labelled, leading to misrepresentation of data (researchers are, after all, only human). At worst, however, entire images may have been manipulated to better fit the hypotheses of the research team.
This type of scientific misconduct has grave consequences for the academic community. The conclusions made in fraudulent papers can trick unsuspecting academics into pouring their time and money into research that is fundamentally doomed to fail. In 2020, a 2006 paper published in Nature (“Lysyl oxidase is essential for hypoxia-induced metastasis”) was retracted after it was reported for containing “several image anomalies”. By this time, the paper had been cited 970 times.
Thanks to websites like PubPeer, the Marples and Poirots of the scientific community are able to bring these suspicious papers to light. Dutch scientist Dr Elizabeth Bik (@MicrobiomDigest) is arguably the most famous image sleuther: under the hashtag #ImageForensics, she shares examples of doctored images with her 90,000 Twitter followers. Like a reverse game of spot-the-difference, she encourages her followers to highlight all of the duplications they can find. The first person to match her findings wins the coveted gold medal emoji.
In some cases, the manipulation is so obvious that it is a wonder the paper passed peer review in the first place. Features within the image are simply – and crudely – pasted over and over again. These features can be anything from peaks in a graph to the bands in a Western blot. Other times, important aspects of the image are deleted from existence. But it is the images that have been painstakingly mirrored, cropped, stretched, and spliced that cause the biggest concern. Spotting these advanced duplications requires a keen eye for detail (and, ideally, an Adobe Photoshop license).
Sadly, it is not hard to understand why some scientists are driven to data fabrication. Publications are the main currency of the science world; the phrase “publish or perish” is well known in the academic community. As Sherlock himself says in A Scandal of Bohemia, “one begins to twist facts to suit theories, instead of theories to suit facts”. It is therefore vitally important that journals safeguard the research record and retract papers suspected of misconduct.
PubPeer allows scientists to comment on papers after they have been published. The website has become a major forum for reporting scientific misconduct, as it gives authors a chance to clear up any misconceptions or justify their findings. Unfortunately, the threat of lawsuits and career repercussions forces many whistleblowers into anonymity.
Just look at the reactions that Bik’s posts can spark. French microbiologist Didier Raoult - whose claims that hydroxychloroquine could cure COVID-19 prompted then-US president Donald Trump to publicly endorse its use - slandered Bik after she reported several discrepancies in his clinical trial data. In a move reminiscent of the former president, Raoult dismissed a study that contradicted his findings (as referenced by Bik) as “fake news”.
But it is through the work of Bik and other image sleuthers that scientific integrity is maintained. While #ImageForensics allows researchers to have fun playing detective, these exercises also help spread awareness among the scientific community. It is important to remember that image anomalies are not always the result of scientific misconduct. Proper interpretation requires good background knowledge of the experiments and equipment used to obtain the data. Nevertheless, the next time you are confronted with a suspicious paper, just think: what would Elizabeth Bik do?
Retraction of Nature paper:
Elizabeth Bik vs Didier Raoult:
From Issue 22: the Dark Side of Science