When "weak links" in science matter -- high profile fraud in Alzheimer's disease research
Trying to raise the alarm about fraud can feel like howling into the wind.
About a year ago I wrote a post trying to gauge how much of the peer-reviewed literature in non-predatory peer-reviewed journals has deliberate fraud. A number that often comes up is 2%. However, according to forensic metascience expert James Heathers, that number is flawed and out of date. According to a recent review article by Heathers, the true number is likely around 14%.
I find many people are either not aware of this issue, or they acknowledge the issue but wave it away as not important.
An example is someone who I met recently who works at the National Science Foundation. From their vantage point, fraud is mostly being committed in China, not in the USA, which they described as having far superior systems for grant dispersal and oversight. This is largely true, but it doesn’t mean fraud from the US is not an important issue.1
Another example is Adam Mastroianni, who has a popular article here on Substack called “Science is a strong-link problem”. Mastroianni conceptualizes scientific knowledge as a series of links. Strong links are added to this chain, while weak links are quickly disregarded. According to Mastroianni, making strong links is the most important thing, much more important than reducing the number of weak links. This is a hopelessly simplistic way of conceptualizing scientific knowledge, but that’s a different story for a different time. I want to focus on the implication of his post for how we should think about fraud. While the post doesn’t mention fraud explicitly, his framework implies we shouldn’t worry much about it. (Also, one section is literally titled, in all caps, “CHEATERS SOMETIMES WIN AND THAT’S OKAY”).
According to Mastroianni, “strong link problems” like science don’t need much oversight or regulation. Mastroianni claims that oversight and regulation can stifle the creative freedom needed to create strong links. As an example of a “strong link problem” he points to music. It doesn’t make sense to crack down on bad music. As an example of a “weak link problem” he points to food safety.2
More generally, Mastroianni and others (like Donald Braben) have a strong bent against oversight on scientific research (in grant funding, peer review) due to a concern that oversight stifles bold thinking and scientific freedom.
However, when it comes to fraud this is not an “either/or” situation. We can fund smart, bold, innovative thinkers and give them a lot of freedom while also having strong, muscular oversight against fraud.
I believe that:
1. Ongoing high-level fraud indicates something is rotten in scientific culture.
2. Fraud often is incredibly destructive.
3. We aren’t doing nearly enough about it.
We need new institutions for oversight that are capable of quickly rooting out fraudulent articles from the scientific literature and leveling harsh penalties on those implicated in the production of fraud.
There are several obvious problems caused by fraud:
1. Wasted resources on the fraud itself.
2. Wasted resources doing replications, follow-up work, etc.
2. Weakened trust in scientific institutions.
4. Poisoning of the scientific literature.
I want to linger on that last point a bit. There are not always “gold standard” textbooks that can be turned to for answers, and even if there are, many do not have them readily available. Increasingly, people are relying on scientific literature published online for answers about important questions. As an example, consider that there are about 20,000 genes, each of which can be mutated many different ways. If someone has a rare genetic disorder, then doctors need to consult the literature to figure out what the effects might be. A similar story exists for the 13,155 drugs listed on DrugBank, and for the ~300,000 known materials. More generally, “reality has a surprising amount of detail”.
It becomes a lot harder to research niche topics when the literature is contaminated with fraud.
Now let’s turn to the main topic of this post:
Recent high-level fraud in Alzheimer’s
It’s easy to get lost in the fascinating details of each case and miss the forest for the trees. So, I’m going to review each case briefly.
Berislav Zlokovic
Number of papers found with suspicious images - “dozens”
Influential papers with fraud - Nature Medicine (2004), Stroke (2013), The Journal of Neuroscience (2013), Frontiers in Neuroscience (2022)
Taxpayer money lost as a result - at least 30 million
Pharma company impacts - ZZ Biotech received 30 million from the NIH to explore drug candidate 3K3A-APC for stroke recovery based on Zlokovic’s work. The drug didn’t work and there were six deaths in the active group compared to one in the placebo group. Suspect papers have been cited in 49 patents by 30 companies, universities, and foundations.
Notes: As explained in Science’s expose, Zlokovic is the top-ranked researcher in the world when it comes to Alzheimer’s and the blood-brain barrier, both in terms of number of papers (25) and number of citations (14,341). Suspect papers affect both the Alzheimer’s and stroke research fields. Two former lab members say that Zlokovic instructed them to make sure their lab notebooks were “clean”, meaning making sure they did not have data and results that challenged their paper’s conclusions.
Eliezer Masliah
Number of papers found with suspicious images - 132
Influential papers with fraud - BMC Neuroscience (2014)
Taxpayer money lost as a result - at least millions (?)
Pharma company impacts - The drug prasinezumab is being developed by Roche Pharmaceuticals largely off of Masliah’s work. So far a Phase 2 trial found no effect. Development of the Parkinson’s drug Minzasolmin was largely based on Masliah’s work. 238 active patents cite Masliah’s fraudulent work.
Notes: Masliah served as Director of the National Institute on Aging’s Division of Neuroscience between 2016 and 2024. NIH conducted an investigation between December 2023 and September 2024, found evidence of misconduct, and notified the HHS Office of Research Integrity of their findings. As explained in Science’s expose, Masliah is easily within the top five most cited researchers in the Alzheimer’s field and within the top ten for Parkinson’s. He is the 2nd most cited researcher for papers on Cerebrolysin for Alzheimer’s, having published 22 papers on the subject. As documented in detail by Mu Wang, research and interest in Cerebrolysin as a treatment for Alzheimer’s and other conditions exploded after Masliah’s fraudulent work, done in conjunction with Ever Pharma.
Marc Tessier-Lavigne
Number of papers found with suspicious images - 12 total, 5 major
Influential papers with fraud - Cell (1999, 903 citations), Science (2001, 743 citations, 2001, 322 citations), Nature (2004, 2009, 834 citations)
Taxpayer money lost as a result - likely millions
Pharma company impacts - Genentech built a lot of their research off of Tessier-Lavigne’s work, who had been an employee there between 2001 and 2008. They lost millions as a result. By 2012, they could not reproduce the results of the fraudulent 2009 work.
Notes: Genentech urged Tessier-Lavigne to retract his 2009 Nature paper in 2012. It took until 2023 (14 years) before it was retracted. It took 22 years to retract two fraudulent Science papers. Concerns about some of his papers were raised several times on PubPeer starting in 2015. People didn’t start paying attention until Theo Baker published in the Stanford Daily in late 2023. While direct evidence linking Tessier-Lavigne to the creation of the doctored images remains elusive, he was forced to resign from the Presidency of Stanford in July 2023. However, he was allowed to continue leading his lab and doing research. According to Elizabeth Bik, the fraud in the Science and Nature papers affects the results, something Stanford investigators disagreed with.
Sylvain Lesné
Number of papers found with suspicious images - 20
Influential papers with fraud - Nature (2006, 3515 citations), The Journal of Neuroscience (2012), Brain (2013)
Taxpayer money lost as a result - at least 10s of millions, possibly 100s of millions lost chasing Aβ*56 as a key culprit in Alzheimer’s.
Pharma company impacts - Cassava Biosciences based their clinical trials of Simufilam partially off Lesné’s work.
Notes: It took 16 years before the fraud was found in the 2006 Nature paper, which is said to have greatly influenced the direction of the field. The main suspect, Sylvain Lesné, remains a professor at the University of Minnesota, Twin Cities.
Hoau-Yan Wang
Number of papers found with suspicious images - 20
Influential papers with fraud - The Journal of Neuroscience (2012)
Taxpayer money lost as a result - at least $16 million, probably a lot more.
Pharma company impacts - Cassava Biosciences based their clinical trials of Simufilam off of Wang’s work. Wang served as an advisor to the company.
Notes: The fraudster remains a professor at CUNY, despite CUNY doing a 50 page investigation that found “highly suggestive” evidence that he is culpable for 14/31 allegations. A June 2024 indictment by the Department of Justice states that he also committed fraud on grant applications between 2017 to 2021. The fraudulent applications led to $16 million in grant funding to Wang from the NIH.
Domenico Praticò
Number of papers found with suspicious images/data - 36
Influential papers with fraud - Molecular Neurobiology (2014), others
Taxpayer money lost as a result - many millions (one grant alone was $3.8 M)
Pharma company impacts - unknown
Notes: Praticò blames a graduate student for flaws in two of the 36 papers. Mu Wang found that some of his water maze data was too good to be true. She wrote a letter about it to the Office of Research Integrity at HHS. Wang, Elizabeth Bik, and others have been urging Temple to investigate Praticò for years, but they have refused to do so. A blog post on For Better Science from 2020 documents a failed whistleblower attempt to get Praticò investigated.
Quotes
“The immediate, obvious damage is wasted NIH funding and wasted thinking in the field because people are using these results as a starting point for their own experiments.” - Prof. Thomas Südhof, Stanford (2022)
“Journals and granting institutions don’t know how to deal with image manipulation. They’re not subjecting images to sophisticated analysis, even though those tools are very widely available. It’s not some magic skill. It’s their job to do the gatekeeping.” - Prof. John Forsayeth, UCSF (2022)
“Dozens of companies are working on anti-Aβ treatment options, all of which may have to undergo a momentous shift if authorities confirm the authors falsified their findings.” - Reed Slater (2022)
“Lesné’s work now appears suspect across his entire publication record. Aβ*56 itself does not seem to exist. Other researchers had failed to find it even in the first years after the 2006 publication, but that did not slow the beta-amyloid-oligomer field down at all. It was going to grow anyway, but it’s for sure that the Aβ*56 stuff turbocharged it, too.” - Derek Lowe (2022)
Addendum
On X, Elizabeth Bik makes an important summary observation:
Further reading
The Academic Culture of Fraud by Ben Landau-Taylor
According to them, in China funding agencies explicitly look at publication count when deciding who to fund and there is a “quota system” — meaning if you are not producing enough papers you won’t get your next round of funding. It appears this has been true for a long time, but the process is now being reformed. Until 2020 Chinese funding agencies offered cash bonuses for publication in top journals. This is an obviously wrong-headed system. People are driven by incentives. Perverse incentives and lack of oversight have created a culture of scientific fraud in China which will take a long time to change.
As an aside, one could argue that the production of scientific knowledge is much closer to food production than music production, because there are specific protocols and techniques that must be followed to produce scientific knowledge, similar to how there are specific recipes that must be followed to produce food. My sense is this is less the case when it comes to music. Doing science right involves walking a very narrow path and constantly working hard to not fool yourself, to paraphrase Feynman. Many specific protocols and methods, collectively called “rigor”, need to be followed carefully to make sure you don’t fool yourself. New researchers need to be taught all of the techniques of rigor and funding agencies have a duty to make sure that the researchers they fund are being rigorous. The importance and centrality of rigor/methods/standards in science is widely agreed upon by scientists, but is a contentious issue among some philosophers of science.
Fascinating analysis. Great points on lost opportunity, costs to public in terms of health, trust, and public funds. Cost to scientific community at large, as well as to individuals who serve on reviews, committees, research time... and lots more. Sad, but very really reflection of the nature of humans. More readily apparent in some areas than others. I would like to think that scientists are more immune to temptations of glory money and power, but the reality is they're not. The question is were to scientists rank in terms of basic malign character relative to other professions, such as politicians, clergy, bankers, investment 'advisers', real estate professionals, car dealers, ....
Disclaimer: I am involved in commercial AD research.
Throat clear: fraud is bad and we should absolutely tar and feather (not to mention fire) the fraudsters.
I want to push back a little and say that I personally think the impact of these frauds on clinical development is overstated.
For example, I do not know of any drugs in development which specifically target Lesné's Aβ*56 species, and yet his fraud is used to cast doubt on the whole Aβ field, including the approved monoclonal therapies (which of course have very small effects and all that, and that's a justified critique, but they're certainly not based on the fraud).
Same for prasinezumab: with or without Masliah, there is still a rationale to at least try and target alpha synuclein.
Cassava is a joke company and a meme stock, they will keep promoting their failed molecule for as long as people give them money. This one is for the SEC to solve, not scientists.
I never saw the "string links" formulation of the problem but I must say I kind of like it as a theoretical framework, as long as we admit that poor quality research should be discouraged and fraud should be punished.
When I was in academia, in an entirely different field, we all knew not to trust the results of certain labs (or colleagues!) without in-house replication. Correcting the record is extremely difficult, as is knowing intent, and not everyone has the persistence to pursue it. I think in practice a lot of shoddy and fraudulent research is just ignored by those in the know, which is like, not great! For sure. But it's workable, and that's the strong links theory in practice.
On the other hand I'm increasingly worried about how this interacts with the training of various knowlege-extracting LLMs which are increasingly used by researchers and often cannot see beyond the context of an abstract.