Voices: Science's editor-in-chief Dr. Holden Thorp on the Stanford president resignation and the detection of research fraud and misconduct.
Earlier this month, Marc Tessier-Lavigne, President of Stanford University, resigned amid mounting pressure stemming from allegations of research misconduct in his lab’s influential research on Alzheimer’s Disease. The move came after a cub reporter at the school’s daily newspaper chased down the story. According to experts, the problems in research from Tessier-Lavigne’s laboratory were neither new nor isolated.
How bad was the misconduct? How do the infractions compare to other recent high-profile cases like the Theranos debacle? How can top-tier medical and science journals detect fabrication and fraud before it’s too late? Can renowned scientists really run major universities and the research labs that got them there?
I asked all of that and more to Dr. Holden Thorp, editor-in-chief of Science. Dr. Thorp wrote a column about this recently, but we had a chance to go deeper in a video Q&A for MedPage Today.
My favorite part of this interview: when Thorp reveals that, yes, editors at prestigious medical journals are now in fact looking at online resources like PubPeer, where everyone from “civilians” to experts can chime in on the quality (and potential problems) found in both published and preprinted academic manuscripts. That’s great news because the more eyes there are on papers, the less likely it is that glaring problems will go unnoticed.
Here’s the audio and full transcript (and the full video is on MedPage Today):
Full text:
Hello, this is Jeremy Faust, MedPage Today. Today, we're going to discuss the allegations of research misconduct that led to Stanford University President resigning this past week.
We're joined by Dr. Holden Thorp. He's the editor-in-chief of Science. Dr. Thorp wrote a column in Science that describes the tension between being a researcher and an administrator. Dr. Thorp, thank you so much for joining us.
Thorp: Thanks, Jeremy. It's great to be here.
Faust: So put this in context for us. How bad is this fraud? Fraud's never good, but how bad are these findings?
Thorp: Well, they compromised some very important and high-profile papers. I guess the one silver lining is that it's been now some time since these papers were talked about on PubPeer, and so people have been talking about the questions here for the last few years. But certainly these were very high-profile papers that had problems and they've been cited a lot, and a lot of people invested in following up on these, so it's not a great example. That's for sure.
Faust: And I guess they didn't find fraud. It's really that something was incorrect and not fixed. Is there some distinction there?
Thorp: Well, I think it's unclear whether they found fraud or not. They didn't find that Dr. Tessier-Levine was personally involved, which is what we normally see in these cases. So I think what happened here is that corners were cut by people in his laboratory and he didn't see that that happened. Unfortunately, we deal with a lot of papers where this is the case.
So, I think that there was pressure to get this stuff written up and to perhaps get the results that they said they had, and people in his lab did things with the images that they shouldn't have done, and he missed it and submitted the papers. What happened from there is pretty well described in the report.
Faust: You of all people can really address this because when you became a Dean and then Chancellor at UNC Chapel Hill, you had to put your science laboratory aside to focus on that. Do you think that it's possible for someone to do science and carry out these academic responsibilities at this level?
Thorp: Look, I think there are a lot of factors here that are being discussed. The pressure that they were under, the pressure to get in journals like Science, Cell, Nature, and a lot of other things.
But one factor certainly was, in my opinion, the number of things that Marc Tessier-Levine was trying to do at the time that these things happened. He was the head of research for Genentech, and he was also running his lab. When I was in a similar situation, I made the difficult decision to shut my lab down.
I see people struggle with this, people who have my job or who have worked for me and gone into these roles. It's hard to give up something that's basically defined you for your whole career. But the students and postdocs in these labs deserve a lot of attention from their PI, and if the PI is running research for Genentech or being a Dean, Chancellor, or Provost at a big-time university, it's very hard for them to get the attention that they deserve.
Faust: And yet, universities love the big name scientists to be the President or the Chancellor. Is there some model that you could think of where it's kind of like President, Prime Minister, King, and Queen? In other words, your David Baltimores can be the people who are the famous scientists leading the organization, and someone who's got a lot of administrative experience can focus on running the place.
Thorp: I think it's possible. Certainly, of all the roles that I had, being a Provost was my favorite because I got to run the university while my boss did a bunch of external stuff.
But I think the problem is that the buck stops here. When you're the President -- there are so many crises that we see in higher education with athletics and student conduct and free speech, and it just goes on and on and on. Now, there's no way to delegate that. Ultimately, people are looking at you.
One of the things that is really unsettling about being the President of university is that sometimes people will say, "The university," and they'll point at you. Like, "I want to know what the university thinks about this." Well, the university has tens of thousands of people, not just one person who's the President. But the way our culture works, it's very hard to push back against that.
So is it possible in theory, I guess, but in practice it's really hard to see how this works productively. There are of course plenty of people who do it, and we don't really know the outcomes for them because they don't get caught up in a matter like this. But as I said in my piece yesterday, there are a lot of examples of people who get these big time roles who have these problems.
Faust: As an editor of one of those top-tier journals, how do you think about this overall question of detecting fraud? Obviously you're not set up for that, but there's PubPeer, which is a website where people put comments, there's just the internet, and we have seen examples of where preprinting might have actually kept fraud from being published in The New England Journal or in Lancet.
Do you think that a journal like science and others should hire the equivalent of hackers to be on the editorial board and make this part of your practice? Or is it still more of a, "We'll look into it when we see a problem"?
Thorp: Yeah. First of all, our reviewers don't spend enough time on papers to pick this up. That's just a fact. A lot of people would say, "Okay, well, it's your fault. Your reviewers didn't catch this." The review process is not really set up to deal with that.
But I do think there are several things that have happened that are improving the situation. Preprints would be at the absolute top of the list. We are very enthusiastic at all times about preprints. Richard Sever, a great colleague, is doing a wonderful job with bioRxiv, and the other pre-print servers are also excellent.
PubPeer is also a great forum, and took us a while, but we watch it carefully for when our papers come up now. That's something we've started doing in the last year or two.
Then we're also getting some software tools that will detect these problems. We're still examining how to implement those, but I would think in the next year or two, you're going to see every top journal with a software approach to this problem as well.
But none of those things existed in 2009 when these papers were published. The real problem that we have, especially with these altered western blots and gels, is that the period between when Photoshop came along and when we really figured out what the rules were going to be was too long and kind of ambiguous.
So, there are a lot of papers that come between the start of Photoshop in the late nineties --when we first started scanning our gels and not just putting the raw photographs in the papers – and about 2010 when it started to become accepted what you could and could not do.
Now, that doesn't excuse what happened in this case. There were clear manipulations that also changed the data, but it was also kind of the Wild West when it came to Photoshop during that time.
Now, we have Elizabeth Bik looking at gels, we have software tools that can look at these, and we have pretty solid rules about what you can and can't do. We've taken a very tough stand against using ChatGPT to generate texts for papers. The reason we've taken such a tough stand is because we don't want to repeat the same thing that happened with Photoshop, right?
So we've got ChatGPT that can write text, some people are saying, "You should use it or you shouldn't use it." What we don't want is 10 years from now, the word sleuths who are doing the same thing with words that Elizabeth Bik has been doing with gels are going to come along and say, "This was written by a machine and there's something wrong with it."
There are a lot of complications to all this, but one of the most important factors is the fact that Science didn't adapt well to the technological advances that came in this case when Photoshop started being used to process gels for publication.
Faust: All right, Dr. Thorp, thank you so much for joining us.
Thorp: Jeremy, good to talk to you.