Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
I just feel there has been so much damage
0:02
done to the credibility of
0:04
scientists. You know, in, particularly
0:06
in the past four years under a president
0:09
who did not seem to support science
0:11
or you know, even suppressed, good information.
0:15
And also the, the power of social media,
0:17
where there seems to be very little incentive
0:20
by the social media platforms to. Bring
0:23
out only the truth and to suppress false
0:26
information. And I can sort of see
0:28
that we want in the United States in particular,
0:30
we want freedom of press and freedom of
0:32
opinion, but I also feel
0:34
that there should be some way of reporting people
0:37
who. Who make false claims,
0:39
like the vaccines have killed more people
0:41
than COVID-19 itself. So
0:51
well thank you for coming on the podcast to listen. Just to get
0:53
started. I wanted to. Kind of
0:55
ask you. The new Yorker just published an article about
0:57
you and in the headline that described you as biology's
1:00
image detective, can you explain what that means? Yeah,
1:03
so, well, I'm a biologist by training.
1:05
I'm a microbiologist and
1:07
I also apparently have some talents to
1:09
find, or to see duplicated
1:11
parts of images or duplicated images.
1:13
So I kind of detect a Photoshop,
1:16
but I can detect if a photo
1:18
has elements that are repetitive. So
1:20
for When we talk about scientific images,
1:22
that could be an image showing
1:25
multiple cells, but all the cells look
1:27
identical. So it appears that those
1:29
have been cloned by Photoshop.
1:31
And so not about biology
1:33
cloning, but by somebody taking
1:35
a cell and stamping that a couple of
1:37
times in a photo. And so that
1:39
is my specialty. I look at photos
1:42
in scientific papers and I will detect
1:44
duplicated elements or just panels
1:46
that have been tested. How big of
1:48
a problem is this well, that is a big problem
1:50
because that means that somebody has. Manipulated
1:53
the results and photos in scientific
1:55
papers are the data. When
1:57
you read a scientific paper, it
1:59
will say we found such. And
2:01
so, you know, this, this
2:03
isn't that experiment and this was the outcome.
2:06
See, figure one. So figure one,
2:08
any figure in a scientific paper usually
2:11
is data that's. It's a photo
2:13
of cells. It's a photo of tissues. Of
2:16
course also be a line graph or a table
2:18
or things like that, but yeah, figures
2:20
are the data. And so if somebody changed
2:23
the results in a photo, or
2:25
if somebody used the same photo to represent
2:27
two different experiments, Might
2:30
be science misconduct, that's data, fabrication
2:33
or falsification. And that's, that's a really
2:35
big, no, no. In science you should
2:37
not be doing that. Is this a
2:39
widespread issue? No, it's
2:41
not. It's it. When you look at my work,
2:43
you might think it's a widespread issue because
2:46
this is what I do, and this is what I find. But
2:48
I actually did a. Research to look
2:50
at how many times I would find
2:52
these duplicated images. And
2:54
so I scanned 20,000 papers that
2:57
had at least one photographic image
3:00
and 4% of those papers
3:02
had a duplicated image or
3:05
duplicated parts within an image. So
3:07
it's about 4% and
3:09
the real percentage of fraud might be
3:11
higher because I'm only looking at photos. And
3:14
so I'm not looking. Tables or
3:16
sequencing data or any other type of data.
3:18
So the real percentage of fraud
3:20
might be higher. I would just estimate
3:22
between five and 10%, but
3:25
I'm only probably detecting the tip of the iceberg
3:28
if you are a really good Photoshop or I wouldn't find
3:30
that. But yeah, so it's 4% of detectable
3:33
duplications in biomedical papers.
3:36
What's kind of the scientist's motivations here. They're
3:39
risking their careers, right. And their professional reputations
3:42
by conduct conducting
3:44
scientific misconduct. And more importantly, there.
3:47
Putting out false, false science and putting a
3:49
bad science. So why are they doing that? Because
3:51
the repercussions are surprisingly
3:54
small, like a lot of scientists
3:57
who do this and who are being caught,
3:59
don't get any punishment for that.
4:01
Like they add best some
4:03
scientists. And this is really. Only
4:06
a very small fraction of scientists who have been
4:08
caught. They might maybe be
4:10
punished by not receiving
4:12
grants for a year or something like that.
4:14
But that's for most scientists
4:17
who already have a lot of brands going on, probably
4:19
not a big problem. So a
4:21
lot of scientists who are being caught doing this
4:23
are still yeah.
4:25
Still having an at, at a university
4:28
and I'm not being fired. That's
4:30
frustrating because it's cheating and
4:32
we would, you would think that the person would be punished
4:34
for that, but there's very, very little punishment.
4:37
And in most cases, people are, yes.
4:39
They'll have a glorious career and get
4:41
more and more. Right. When
4:43
did, when did this first come across your radar as
4:45
something you were interested in and an issue that you saw
4:47
was significant? I
4:50
started his work working on plagiarism,
4:52
actually. So I heard on a podcast or
4:54
I was reading about science. Misconducts,
4:56
but specifically about plagiarism. And
4:59
I thought let's just check a random
5:01
sentence that I had written in a science paper
5:04
in in Google scholar. So I put it, put
5:06
that sentence between quotes in Google scholar
5:08
expecting only to find my own paper. And
5:11
I found another paper published. Predatory
5:15
publishing book or like some, yeah,
5:17
some, some strange online book that was
5:19
free for download, but it was actually my texts.
5:21
So they had used my sentence and
5:24
pass it off as their own in this paper.
5:26
And it turned out that this paper, not
5:28
only I'd use my sentence, but the sentences
5:30
of many other scientists. So it was sort of this,
5:32
this patchwork of many different scientists
5:35
at different, many different papers that they. Put
5:37
together and sort of passed off as a
5:39
new paper, but it was all plagiarized
5:42
text. So I, I worked
5:44
on plagiarism for about a
5:46
year. And then by another
5:48
coincidence I spotted in a PhD
5:50
thesis, a Western
5:52
blots. It's a protein bar lot. It's a photo.
5:54
And I, it had a very particular smear
5:57
that are recognized. And then I saw
5:59
a couple of pages in another chapter
6:01
or so in that sense, PhD
6:04
thesis. I spotted the same photo, but
6:06
it was upside down and it had been used
6:08
to represent the different experiments. And,
6:11
but yeah, I recognize it. It had this weird
6:13
little spot or smear, so
6:15
that was not good. And this paper had been published
6:18
in a scientific paper as well. And
6:21
I recognized I had some talent to, to do that.
6:23
So it was all by coincidence mainly, but
6:25
it. It's one of those moments
6:27
that sort of make your career or make
6:29
your career change? Yeah, if I hadn't
6:31
seen that, then I probably would never
6:34
have worked on this while you were doing
6:36
this. What was your day job? I
6:38
worked at Stanford, so I was a microbiologist.
6:41
I worked on the microbiome
6:44
of Marine mammals and humans. So the microbiome.
6:47
The bacteria that live inside our bodies
6:50
and, or on our bodies on our skin. And
6:52
I was working on, on the microbiome of
6:54
humans, but also dolphins and
6:56
sea lions. And that was my day job. I
6:58
was, I guess, a regular scientist
7:01
working at Stanford and writing
7:03
papers, doing research. And I was
7:05
doing this image duplication
7:07
searches in the evenings or in the weekend. So it
7:09
was sort of my, my hobby. I
7:12
yeah, I saw, well, I think you're, you're being a little
7:14
humble here. Cause I saw on Google scholar that you have
7:16
a paper that has like 20 K citations,
7:18
something crazy like that. And I, I
7:21
worked in a PhD lab in college and so
7:23
I knew a bunch of PhD students here and PhD
7:25
is you're by far the most cited person. I know there's,
7:30
there's plenty of other people who will play. I'm
7:32
just a very modest scientist by their standards.
7:34
So there's always, people have published more,
7:37
I think you know, four, four at the point. Of
7:40
my career that I'm at, I
7:42
am a probably, yeah,
7:44
sort of a middle of the pack type of scientist,
7:47
but yes, there's one paper that we published
7:49
in science and I'm the second author of
7:51
that. And that paper has a lot of
7:53
a lot of citations is what was one of the first
7:56
publications. Analyzing the microbiome
7:59
of humans using DNA sequencing.
8:01
So there had been other papers. We were not the first,
8:03
but it was one of the first large-scale papers.
8:06
And that has been cited a crazy amount
8:08
of times, I guess it was published in science.
8:10
And so I was incredibly
8:13
lucky to have worked on, on that project.
8:15
And yeah, I think the, the
8:17
paper still stands as we, we, we made
8:19
sure it was high quality and. No
8:23
image duplication. Oh, wow.
8:25
There's actually no phone auto in it. So like
8:27
most scientific papers. There's actually no photo in
8:29
this. Just, just line graphs. And
8:32
but yeah, I can vouch for
8:34
it that there's no, there's no science misconduct.
8:36
I'm sure there's errors in it. Like in any paper
8:39
that, in which you analyze, you know, thousands
8:41
and thousands of, of DNA sequences, there's.
8:44
It's very hard to not make any errors. We all make
8:46
errors. But it's it's done with the best of
8:48
intentions and it has stood up
8:50
to the, to the test of time. It's still being
8:52
cited. So Elizabeth
8:55
in do some research for this pod, or this
8:57
particular subject, I realized that
8:59
there are a number of ways to publish
9:01
fraudulent data in science. The one
9:04
that you specialize in, which is image doctoring it,
9:06
I guess it's somewhat prolific, like what you
9:08
just described. But then there are other ways like people
9:10
will run experiments. Nine out of 10 times
9:12
the result is no, but then like one out of 10 times,
9:15
it ends up being exactly what they want and then they get
9:17
that published. So what's your sort of take
9:19
on the full wide variety
9:21
of flawed in science and whether the incentive
9:24
structures that allow image doctoring to happen or
9:26
the same instead of structures that allow this to happen.
9:28
How would you break down sort of the incentives
9:30
that, that lead to both? Well,
9:33
so the incentives in science are,
9:36
are to publish. So we, as scientists
9:38
are. Encouraged, but also
9:40
almost forced to publish because it's needed
9:42
for our careers, like as a postdoc
9:45
or a professor you need to, or you're expected
9:48
to publish an X amount of papers per year.
9:50
And unfortunately, scientific publishing focuses
9:52
on positive results. So
9:55
if you have done a lung study
9:57
showing that a particular drug does not help
9:59
against the particular cancer, that's
10:01
not a very publishable. Paper,
10:04
because it's sort of a negative result. It
10:06
shouldn't be, but unfortunately, a lot of journalists
10:08
will say, well, that's result is not very
10:10
novel or, you know, earth shattering.
10:13
We want to have a positive results. So
10:15
the incentive to publish positive
10:17
results is one of them. Important. Yeah.
10:19
Incentives to cheat because people want
10:22
to have a positive results. And like you said, if
10:24
you have you know, 10%
10:26
only 10% of your experiments gives
10:28
the results you would like, then that's the
10:31
experiment that you'll pick for your paper. So
10:33
that's called cherry picking is basically
10:35
picking the results that you
10:38
like to see that fits your own hypothesis. But
10:40
ignoring the results. They do not fit your
10:42
hypothesis. And that would be
10:45
also called publication bias or like
10:47
we are, we're all biased. We all want our
10:49
experiments to work out a certain way. And if
10:51
it doesn't do we accept those results
10:53
or do we keep on trying until we have a positive
10:55
results? And, and that's still a big step
10:58
to where science misconduct. I do feel
11:00
that it's, it's cheating in a way, but
11:02
it's. I feel as bad
11:04
as really faking or forging
11:07
results. Like when you have have
11:09
you met, if you have measured a couple of things
11:12
and you change the results, you've changed the
11:14
values so that they cross a particular
11:16
threshold and suddenly your negative sample becomes
11:18
a positive. That is where
11:21
we're really talking about science misconduct. So
11:23
there's, there's a whole range. Steps
11:26
in between from publication
11:28
bias, towards P hacking, which
11:30
is sort of the cherry picking where you keep on
11:32
doing statistical tests. And
11:34
there's one statistical test in which your results
11:37
are only is significant. You pick that. Really
11:40
changing or fabricating results
11:43
that which so fabricating
11:46
and falsification, those are considered science
11:48
misconduct to get to with plagiarism, by
11:50
the definition of the office of research
11:52
integrity, she hacking and
11:54
publication bias are not necessarily
11:57
included in the pure
11:59
definition of science Smiths for that, but there's a
12:01
lot of gray in between. There's, there's a,
12:03
it's hard to draw the line. What is misconduct
12:06
and what is. Bias. Right.
12:08
I think in reading the New York article specifically,
12:11
I was kind of surprised to find that you had
12:13
found issues and some of the most
12:15
important and prestigious journals
12:17
and articles. And when I'm always listening to
12:19
news or other podcasts
12:21
about science, they're always referring to peer reviewed
12:23
articles. Or these editors at these
12:25
journals have the highest standard. But
12:28
obviously not because, because you've found you've,
12:30
you've discovered some sort of fraud. What are these editors
12:32
missing? Like what is, what is the issue within
12:35
this organization that allows stuff
12:37
like this to get published? Oh,
12:39
there could be all kinds of issues that editors
12:41
are either not paying attention to, or just
12:44
not trained to find problems with papers.
12:46
So you would hope that an editor would
12:48
find them. Yeah. Obvious Photoshops
12:51
in, in, or most obvious errors
12:53
in papers, but an editor
12:55
is usually a person who's unpaid
12:57
who does editing sort of as a side
13:00
job, but might be a very busy professor
13:02
who being asked to do an editing. B being editor
13:04
of a journal. So basically
13:06
what they do is they, they got the manuscripts
13:09
that are usually pre-screened and
13:11
then did need to find peer reviewers, or
13:13
also busy and unpaid and, and
13:15
have no really no time to really look
13:17
carefully at a paper. And
13:19
and then when they received the peer reviews, they
13:21
sort of compile that and
13:23
make a final decision. But very often
13:26
I've been an editor myself for a very short
13:28
amount of time. I found it very hard. It's
13:30
You don't really have time to read the paper yourself.
13:32
You sort of rely on your peer reviewers to do
13:34
a good job. And sometimes they also do don't
13:36
do a good job. So it's, it's really tough because
13:39
none of these jobs are paid. And
13:41
unfortunately we have to pay the publisher a lot
13:43
of money to get our work either depending
13:46
on the model of publishing to, to
13:48
get the paper published. So where that money
13:50
goes into is, is not very clear.
13:52
Everything is published online. So it's.
13:56
And, and also editors are not trained to
13:58
find these problems. I'm looking at it with
14:00
a lot of experience. I
14:02
I've seen a lot of different types of Photoshops
14:05
or photo duplications. And so
14:07
I'm very trained for these situations
14:09
because I've seen so many of them, but
14:11
a lot of people don't really see these problems
14:13
until you pointed out to them. I point
14:16
point I do a lot of puzzles on
14:18
Twitter. I will post them on their image forensics.
14:21
And then of course, when I write. One of
14:23
those images, people know there's something wrong
14:25
and they'll usually find it. But
14:27
people forget that I've seen hundreds
14:29
of images maybe before this one
14:31
with the duplication. And so once,
14:33
you know, there's an arrow, you'll find it. But if you
14:36
just quickly look at fairs, you
14:38
might not spot it. And you need to be
14:40
told that this might be a thing before you start
14:42
to see that yourself. So if you were
14:44
a benevolent dictator, what would you change about
14:46
the system to avoid these issues? It's
14:48
really hard because fraudsters are going to fraud.
14:51
And the only thing you can
14:53
do is, is to like
14:55
ask people to ascend
14:57
in raw data. That would be sort of
14:59
an extra hurdle, but if a real fraudster
15:02
wants to fraud, they're going to fraud. So I,
15:04
I have no illusions that. Make
15:07
science a hundred percent foolproof and fraud
15:09
proof. There's always going to be people
15:11
who want to cheat the system because
15:13
we, we put so much emphasis
15:15
on outputs on science papers as,
15:17
as the output of scientists. And
15:20
it's only when we have replicated
15:23
a paper that we sort of know it was
15:25
probably true, but you can never a hundred
15:27
percent be certain that
15:29
everything in the paper was honest and
15:31
that's very unfortunate. Science
15:34
in a way is, is about finding the truth.
15:37
I've always felt like when you are in
15:39
science, you want to discover
15:41
a particular pathway or a
15:43
particular bacteria, or
15:46
you want to discover what is true.
15:48
What is, what is the truth about
15:50
a particular biology process?
15:52
And so I've always felt that science should be
15:54
about reporting the truth. So
15:57
for scientists to fraud, I feel that's
15:59
a huge violation of, of our profession
16:02
as well. Yeah, I definitely
16:04
agree with that. So I'm, I'm someone who reads
16:06
a lot of papers. Anytime. I'm curious about something I'll
16:08
hop on Google scholar and I'll, I'll see what I can find.
16:10
But I I'm, I'm not a PhD.
16:13
I don't have that much training in this field.
16:15
And I have a lot of friends who do the same thing as me and don't
16:17
have any experience with this. how do we read
16:19
papers and say, this is something which we should, we should have high
16:21
degree of credibility. And this is a really good paper versus
16:23
this is something that, you know, maybe we shouldn't put
16:26
too much confidence in. Yeah,
16:29
that's a good question. I don't have a standard
16:31
answer for you because even papers
16:33
that have been published in high impact journals
16:36
by. You know, officers who
16:39
work at the institutions that seem to
16:41
have some credibility, even
16:43
those papers have been caught with
16:45
fraudulent data. So it's not a hundred
16:47
percent guarantee, but having said that
16:50
papers that have been published in science
16:52
or the Lancet, usually with
16:54
some very big exceptions, usually
16:56
are more credible than papers in that are,
16:58
for example, published on a preprint server
17:01
that have not been peer reviewed. Those are the two
17:03
extremes, but yeah. There,
17:06
there has been a big paper in published
17:08
in the lenses that had been busted
17:10
retracted last year, because it was based
17:12
on probably based on fraudulent data. And
17:15
so that's one of those big exceptions that
17:17
makes headlines and that make a lot of people
17:20
who are not scientists, think that all
17:22
science is flawed. Well, that was really
17:24
the exception. It's it's like saying,
17:26
yeah, thoroughness, you know, it was a company that
17:29
did not really do well, so
17:31
we kind of trust any biotech company.
17:33
Anyway, you cannot make those, those
17:36
extra population extrapolations
17:38
based on one bad apple. It's
17:40
it's usually the. Those cases
17:42
make headlines and for good reasons.
17:45
And that was a fraudulent paper by at
17:47
least from old evidence I've seen, but
17:49
it was hard to recognize it as a fraught Olympic
17:51
for, I did not recognize if myself, either.
17:54
I actually tweeted about this paper, my
17:56
haters, my trolls who are my
17:59
loyal enemies on Twitter are still
18:02
saying, oh, big tweeted about
18:04
this paper. So she cannot detect any
18:06
fraud. Right. And it was hard just
18:08
to look at that paper and realize it wasn't fraud.
18:10
You really had to dive deep into the paper,
18:12
knew a lot about particular numbers
18:14
that were misreported to find out
18:17
that that was fraud. So it happens anywhere. Those
18:19
cases make big headlines, but in the end, Usually
18:23
you can trust those, those journals. But yeah,
18:26
it's a, there are exceptions, of course. So,
18:28
so let me, let me play a sort
18:30
of devil's advocate really quickly, or at least
18:32
from what I've read and what I understand by be totally
18:35
off. I came across these pre-publication
18:37
sites, right? Like I haven't
18:39
written down here AR XIV and
18:41
bio R X archive
18:45
archive, bio archive, archive.
18:49
That's how you pronounce it. Okay.
18:51
Sorry. This is So
18:56
not everybody knows where I have that on this argue.
18:59
Right. And the argument that I basically read is, well, sometimes
19:02
it's worthwhile to publish some sort
19:04
of science output. Just get the output
19:06
out there, even if it's just an idea, even if it isn't
19:08
peer reviewed, even if it isn't a hundred percent
19:10
accurate and has veracity. Just to get that idea out
19:12
there, you know, into the, into the minds of people that
19:15
might do more research and build on it, even
19:17
though it's like incredibly low barrier to entry and
19:19
anybody can get it out there. Is that generally
19:21
a good thing for science? Do you think? I
19:23
believe so. And especially in the case where
19:26
we were last year at the beginning
19:28
of an epidemic where. Quite
19:31
frankly, we're all in a state of panic where
19:33
that was, you know, a lot of mortality,
19:36
a lot of people dying, a lot of people getting
19:38
sick, a new virus, nobody really knew
19:40
about, you know, th the new enemy
19:42
was in, in a situation like that. We need
19:44
science to be fast, and we need
19:47
to have a very quick model
19:49
of scientific publishing. So if a person
19:51
has found a result
19:54
that is worth sharing, that might save lives.
19:57
There's a big argument to make, to
19:59
publish this quickly, even though it might mean
20:01
publishing before peer review, but
20:03
just getting it out there so that a lot of people can
20:06
read it and, and benefit from
20:08
these results. But there's a delicate
20:11
balance between wanting to
20:13
publish fast and doing good science. So
20:15
those things are. Yeah, they're, they're, they're
20:17
two ends of the spectrum. It's, it's two parts
20:19
that are usually not in agreement
20:22
with each other because science, if it's
20:24
done well, it's very slow. It's
20:26
a Spain awfully slow. It's like looking for
20:28
tiny details, having long
20:30
arguments with other scientists about how
20:33
to interpret the particular results. Yeah,
20:35
that just is not, cannot be done
20:37
in a very fast way. And so it,
20:40
it's, it's finding this balance between
20:42
publishing results really fast, but
20:44
knowing it's on a preprint server,
20:47
it's not being peer reviewed. It's
20:49
just a view of one particular lab. And
20:51
that couldn't be very right biased because no
20:53
other people have had a chance yet to
20:55
carefully digest it and
20:57
give feedback and go through these normal
21:00
and slow processes. So it's. It's
21:03
I'm all for pre-print service,
21:05
but it comes with a lot of caveats. You
21:07
need to interpret it as just
21:10
a fuse of one lab, not being peer reviewed
21:12
and take it with a grain. So
21:14
one high profile instance that
21:17
I think most people are familiar with of
21:19
of a paper being just rushed out before
21:22
it was ready. Was the, was
21:24
Trump's favorite? Hydro, hydro
21:26
chlorine. Sorry. How do you say that? Hydroxy
21:30
chloroquine study. That's just like God got
21:33
shuttled out and I believe you were one of
21:35
the early scientists to say, this
21:37
is, this is bad research. Right? So
21:39
can you, can you kind of talk about that situation? Sure.
21:42
So this was a paper by the
21:44
group of professor Howell in Maaseiah
21:47
in France. And he claimed that hydroxy
21:49
cork Quinn was a
21:51
really good medication to get
21:53
rid of the virus. So he looked at patients
21:56
who had the virus who were positive for the PCR,
21:58
and he looked at clearance of the virus. Repeatedly
22:01
testing these patients and seeing when
22:03
they would become negative. And he showed
22:06
in his paper, which was only, I believe 40
22:08
patients. So it's a very small study and
22:10
he had three different treatment groups. So some
22:12
people were not treated. Some people only got hydroxy
22:15
Clarke when and the third group got hydroxychloroquine.
22:18
Plus I see assay throw mycin, which
22:20
is an antibiotic. And he showed that
22:22
the the both groups that had the hydroxy
22:25
Clarke. Treatment that those people
22:27
cleared the virus. So got PCR negative
22:30
faster than the people who did not
22:32
receive any of those drugs,
22:35
but the, the groups were really small. So if you
22:37
have 40 patients and you divide them over three groups,
22:39
you can already see that the numbers get pretty small,
22:42
but there are a lot of flaws
22:44
with this study. So one of the things was that. There
22:47
were six patients who were in
22:49
the hydroxychloroquine groups in either one
22:51
of these groups who didn't most of these
22:53
patients were, did not really do very
22:55
well on the hydroxychloroquine and they were
22:57
left out of the study. So they started
23:00
with a particular patient group
23:02
that's but six patients were left out. So one
23:04
of them died. Three, I think two
23:06
of them got really bad to got really
23:09
sick. So they were transferred to the intensive
23:11
care. One patient got little. Side
23:13
effects are two patients and one
23:15
patient just walked out of the study. So it wasn't
23:18
pretty clear. So it looked like the
23:20
researchers might have decided
23:23
to do this cherry picking that we talked about
23:25
previously, where, you know, the results
23:27
were really quite what they had hoped. You know,
23:30
if one patient dies on your drug, You
23:32
should not leave them out of the study. Right? science,
23:37
it's us. So basic, like just
23:39
say, I did not want this result. I'm just going
23:41
to leave it out. And that was actually, it had
23:43
noted that they had written it in the paper.
23:46
So who else might to have
23:48
left? Left? That was just one,
23:50
there were many other problems with this paper. So I
23:52
wrote a long review about it. There
23:54
was problems with ethical, ethical,
23:57
the dates of approval, or first the start of the
23:59
study that there was some problems
24:01
there. He included some children,
24:04
even though he wrote that he didn't think Lew children,
24:06
but he did. And then. There were differences
24:08
between the treatment group. So those people
24:10
were all treated at a different hospital than
24:13
most people who were not treated. And so there were
24:15
all these differences between the different treatment
24:18
groups that you would not expect. Red
24:21
or rigorously set
24:23
up scientific study, you would do shoots,
24:25
randomize your patients. And he
24:27
didn't do that. He appeared to have handpicked patients
24:30
and maybe those patients who were
24:32
on hydroxychloroquine were already
24:34
less sick to start with, or
24:36
maybe farther in their in their disease
24:38
status. So they would have cleared the virus even faster.
24:41
So all kinds of problems. And I,
24:43
I wrote a, a critical blog post about
24:45
that, and that got me into trouble. Can
24:48
you elaborate on that? Yeah. So
24:51
obviously professor Al
24:53
did not enjoy my critique
24:56
and and I can understand that I
24:58
understand that he was not happy with my critique
25:00
and I, so I raised the concerns
25:03
not only by writing about this on a
25:05
blog post, but I also posted on a
25:07
website called pub peer, which
25:09
is a website where you can leave comments on scientific
25:11
papers. And he did
25:14
not answer to my command. Stare
25:16
in study, started calling me all these names.
25:18
So he's called me a witch hunter can
25:20
sing play, which means like a crazy woman.
25:23
He called me a girl who hunted me down
25:25
and, and no, all kinds of, not
25:28
very nice words, but yeah.
25:30
So, and he also some
25:32
of his people who works for him who
25:34
are working for him in the same institution,
25:37
Started to harass me on Twitter. And so
25:39
there was one one of these professors
25:42
like shabby and started to ask me all
25:44
these questions on Twitter. Which
25:46
may boil down to who are who's
25:48
paying you, are you being paid by
25:50
big pharma to bring down
25:52
my professor, professor And
25:55
in the meantime, I started to look at more
25:57
papers by that. Well, and found more problems.
25:59
So there were some, actually some, some
26:01
image problems. So some image, duplication
26:03
problems, but also other problems with
26:06
ethical approval of some of
26:08
his studies. So it appeared that there's not
26:10
just this hydroxychloroquine paper that had
26:12
some. But also a bunch
26:14
of other papers. So I posted all of these on
26:16
puffier and other people started chiming
26:19
in finding more and more problems. So by
26:21
now I think this professor has
26:23
270 papers
26:25
from his group that are, have been flagged
26:27
on pop here. And he's becoming a bit
26:30
annoyed with all of us and yeah, he has
26:32
now. Threaten me with a lawsuit. So he
26:34
has filed a complaint against me with
26:36
the prosecutor in Maseo
26:38
claiming that I harassed him and that
26:41
I extorted him and blackmailed him.
26:43
And that's all based on two answers.
26:45
I gave them on Twitter where they asked me,
26:48
who's paying you. And I said, oh yeah, You
26:50
can donate money, I'll make patron accounts.
26:53
And another one I said, well, I can, I'm a consultant.
26:55
And so I could check papers if you want,
26:57
if you want me to check some papers, happy
27:00
to do so, as long as you pay me. So
27:02
he claims that's blackmailing. I
27:04
kind of imagined that's blackmailing, but yeah, it's
27:07
you filed a police report with
27:09
the prosecutor in the Messiah and
27:12
this case is under investigation. And
27:15
I, I hope this will not lead to a
27:17
lawsuit because I don't
27:19
think I did anything wrong, but I'm
27:21
not quite sure how the legal system
27:23
and friends work. So, so for
27:25
now it seems to be. Threats to try
27:27
to silence me, but I've already sat on Twitter
27:29
a couple of times. I'm not going to be silenced. I'll
27:31
keep on a stand behind all my
27:33
questions. You can answer them on pop here.
27:36
And I don't think that a scientist
27:38
should be resorting to legal
27:40
steps to silence your critic
27:43
hustlers. But yeah, I guess that's. No,
27:45
a couple of other authors in
27:48
some cases have, but not all himself
27:50
and should have. Yeah, I have not answered any of the questions.
27:54
I saw a petition. I think it
27:56
was like a thousand scientists who came out to support you.
27:59
Why? I mean, how could you not, this is, I
28:01
don't know, this, this is just so outrageous is unbelievable
28:04
that like, not only is this
28:06
guy putting out garbage science that
28:09
has affected us that has affected the United States
28:11
because the president used that as,
28:13
as policy. But he's also going after. Whistleblowers
28:16
who are trying to keep science clean. This is so outrageous.
28:19
It is. Yeah. And unfortunately, that's,
28:22
I'm not the only person who has
28:24
been harassed or threatened on
28:26
Twitter or even in real life.
28:28
I have not been to that country in real life, but there are
28:30
a couple of scientists who are
28:33
just trying to bring out good news
28:35
or, well, let's say honest news
28:37
and try to. Go against
28:39
people who spread misinformation. There's
28:42
so much misinformation right now on social
28:44
media, where, for example, there's,
28:46
there's these tweets where people claimed that more
28:48
people have died of the COVID vaccine
28:50
than of COVID itself. And as a
28:52
scientist, you kind of be silenced. You kind of
28:54
look at these numbers and
28:57
just looked out away because it's completely not
28:59
true. And so a lot of scientists will say,
29:01
that's not true. We'll get to here, here at a number. But
29:04
then there's all these people are not scientists,
29:06
usually who claim they know better.
29:08
And they have fun to another website that disagrees
29:11
with all these hundreds of scientists. And so you get
29:13
all these very polarized
29:15
situations where scientists try
29:17
to bring out honest
29:19
information and based on facts and
29:21
other people say that the
29:23
facts are. Yeah, all these, these
29:26
things, these wars going on on Twitter,
29:28
and sometimes scientists have been threatened.
29:31
And there's actually one scientist in Belgium
29:34
who is now living under police protection
29:36
in a secret location because
29:39
he's being threatened. Then I was the soldier
29:42
who has escaped from
29:44
the army or something with a little of weapons
29:46
and it was trying to kill the scientist.
29:48
And it's just very strange
29:50
situations. We'll see that as well. Yeah,
29:53
I know there's just this like rising tide
29:55
of scientific misinformation. And I can think of one
29:57
big reason why. But of course, social media
29:59
has also, I think I contributed to this
30:01
now. Just anyone can publish without any
30:04
sort of. I dunno. I dunno, peer
30:06
review is not the right term for when people who are not
30:08
scientists say stuff, it's like the idea of sharing an
30:10
article without reading it first. Something like that.
30:12
Elizabeth, I think on this topic. And this is something
30:15
I've been thinking about a lot, is that it
30:17
just feels like in the since the pandemic, I guess
30:19
it's not exactly what we're, but since it we're coming to an end
30:21
to it it seems like the public. Really
30:24
doesn't trust the scientific community or it's
30:26
at an all time low. What do you think can
30:28
be done to improve
30:30
that trust whether it's in the United States or
30:32
around? Oh,
30:34
that's a great question. I. I
30:37
don't really know to answer because I,
30:40
I just feel there has been so much damage
30:42
done to the credibility of
30:44
scientists. You know, in, particularly
30:47
in the past four years under a president
30:49
who did not seem to support science
30:52
or you know, even suppressed, good information.
30:55
And also the, the power of social media,
30:57
where there seems to be very little incentive
31:01
by the social media platforms to. Bring
31:04
out only the truth and to suppress false
31:06
information. And I can sort of see
31:08
that we want in the United States in particular,
31:11
we want freedom of press and freedom of
31:13
opinion, but I also feel
31:15
that there should be some way of reporting people
31:17
who. Who make false claims,
31:19
like the vaccines have killed more people
31:22
than, than COVID-19 itself.
31:24
And, but yeah.
31:26
Then other people will say, well, if you suppress
31:28
that opinion, then that's oppressing
31:31
freedom of speech. And I, yeah, I,
31:33
I think that's a very hard
31:35
to solve. Issue. And I,
31:37
I sort of want this country to be
31:39
about freedom of speech, but when
31:42
that turns into misinformation, that
31:44
could actually cost life. I do feel there needs
31:46
to be drawn a line somewhere. And
31:48
as scientists, we are all very frustrated
31:51
that there's no way to report on
31:53
Twitter, false information. There is
31:55
actually no bottom. There's no way to report
31:57
people who send me emails saying
31:59
you belong in jail, or you are
32:02
a fraud. Like I cannot report
32:04
that I've reported several of these tweets and
32:06
I always get to hear from Twitter.
32:08
We don't feel that violates our rules.
32:10
Like you can actually say a lot of
32:13
things to each other before, before
32:16
tweets are being taken down. And I,
32:18
yeah, it's, it's this delicate balance
32:20
between freedom of speech. Yeah. Still
32:23
trying to be polite to each other, and I'm
32:26
not sure how to solve this, this this
32:28
is a very important question with with
32:30
a lot of aspects to it and yeah, just
32:32
don't have the answer. Do
32:34
you do you think all scientists, whether
32:36
it's physics, math, biology,
32:38
geology, whatever have a problem with
32:40
false data or is it just a bigger issue
32:43
within certain subsets yeah, so
32:45
I feel fraud in science is
32:48
probably anywhere in any particular
32:50
field of science. I focus
32:52
on images, which are a part
32:54
of molecular biology type of papers
32:56
because they have a lot of protein blots or DNA
32:59
blots. And so those are generally
33:01
photos, but there's also.
33:03
A lot of other types of data,
33:05
like optical spectra, where I found
33:07
fraud in I haven't really looked into
33:10
a lot of other fields, but I do feel there's
33:12
probably fraud everywhere, but I don't know
33:14
enough for example, to, to look
33:16
at the mass paper or a geology
33:19
paper to find a potential problem in it,
33:21
because that's not really. My
33:23
background, I, I look at these papers
33:25
and I just see numbers or graphs,
33:27
and I just don't understand what they mean. So I
33:29
kind of detect problems in them, but I'm
33:32
pretty sure that fraud is everywhere, but
33:34
I also think it is important to
33:36
be it's, it's easy to listen to my
33:38
story and the story of
33:40
misinformation and scientists and, and I want
33:43
to make sure that we. Confuse these two things
33:45
because there's fraud in science and that's what I
33:47
work on, but I also want to make
33:49
sure that there's, that most science is
33:51
to be to be trusted. And I feel
33:53
it's very easy to hear my steel, sorry.
33:56
And interpret like, oh, all science is,
33:58
is flawed. And we kind of trust that. And
34:00
at the same time I'm telling no, we should trust
34:02
science. And I, I feel that's a very
34:04
important thing to To distinguish
34:06
between these two things. So I, I will
34:09
say that there is fraud in science. It's probably
34:11
everywhere. There's fraud everywhere. There's fraud in banking,
34:13
in construction. You know, what
34:16
is there, there's probably no field that
34:18
you can think of that has no fraud. So science is
34:20
not immune to that either, but as
34:22
a whole, science is about finding that truth.
34:25
And and it's the only solution we have. I
34:27
feel to solve the big problems that we're
34:29
currently facing in the world. Epidemics
34:32
and climate change and,
34:34
and things like that. And I think by now, most
34:36
people will be convinced that for example,
34:38
the earth is not flat. And I feel that
34:41
a lot of these misinformations in science
34:43
are based on, you know, the earth is
34:45
flat. Data, like there's, there's no real data
34:47
to believe that that's the case, but people, if
34:49
they want to believe that they'll believe in that.
34:51
Right? The reason we ha the reason we live
34:53
longer than 40 years and we have cell phones
34:56
and the internet is because of science. Exactly.
35:00
I, I yeah, I I'm, I'm like, at this point
35:02
I'm like, we need more Elizabeths in the science community, but
35:05
people like me, I'm not, I'm really
35:07
not the only person, but the most of these people
35:09
work anonymously for good reasons,
35:11
as you can see, because I'm being haunted down
35:13
by the French you know,
35:15
disinformation, trolls so most people
35:17
will choose to, to do this work
35:20
anonymously, but I'm definitely not the only person
35:22
doing this topic. Do you
35:24
think that in the near future, we'll
35:26
be able to train an algorithm to spot
35:28
the doctorate image if given like 10,000
35:30
images, basically give your eye your
35:33
particular unique talent to an algorithm.
35:35
Is that a possibility? Yes. And I actually, I'm
35:37
going to take a sip of
35:39
water because I turned this into a drinking game
35:41
because I get this question so many times
35:43
on Twitter. So I'm just going to take a sip
35:45
of water. Delicious.
35:49
So a lot of people will
35:51
say, oh, I can, I can ride
35:53
it tool on a Friday afternoon. That can do what
35:56
you can do. It's much harder than
35:58
you than you might think, because a lot
36:00
of these duplications
36:02
are not pixel to pixel identity. So
36:04
science images have usually
36:06
been compressed a lot. They're inserted
36:08
in a PDF. There's all kinds of image
36:11
compression and data processing
36:14
that. Made one image
36:17
looked like an outer, but not image, not
36:19
pixel to pixel. So you're kind of. Do
36:21
your standard pixel to pixel comparison
36:23
and find these things. A lot of them, a lot of
36:25
the times the images also rotate or zoom
36:28
dinners ripped out or like mirror.
36:30
So it's a little bit more complex than that.
36:32
And I've actually participated in a DARPA
36:34
challenge where I came with my data stack
36:36
of flawed images and good images. And
36:39
nobody could crack the code. There were several groups
36:41
that, that all claim that
36:43
they could ride on a Friday afternoon. They could write
36:46
this probe into detectives. And we're now
36:48
three years later and now they're starting to
36:50
develop tools that can actually do
36:52
this. So it's, it's pretty hard. But
36:55
on the other hand, yes, this is information. This
36:57
technology will be there and it's,
36:59
it's actually getting ready. There's a couple of tools
37:01
I'm already starting to use that
37:04
are. Starting to find applications,
37:06
but, and in some cases they're better
37:09
than what I can find. They're definitely faster,
37:11
but there's also duplications that I just see
37:14
what my, just my eyes
37:16
and the suffer just cannot see it.
37:18
I'm like, come on. This is there. It's so clear.
37:20
Yeah, so it's, we're still, we still have a long
37:22
way to go and it always needs human
37:25
interpretations. Software
37:27
can easily, maybe in the end pick up
37:29
a duplicated image, but in some cases
37:31
there are duplications that are expected
37:34
and actually quite normal. Where,
37:36
for example, you do a control experiments and
37:39
you compare to two particular drug
37:41
treatment, but then later you have the same photo
37:43
of the control experiment, and you compare to another
37:45
experiment. So in those cases you might see the
37:47
same photo. But it's a total,
37:50
a totally normal and acceptable
37:53
way of, of reusing the photo. So
37:55
the software might still detect it as a duplicate,
37:58
but then you need a human to interpret. Yeah.
38:00
Well, this is actually the same experiment, so that's fine.
38:02
The advantage of software will be
38:04
in the end that any image in a
38:06
manuscript that is sent into
38:09
a journal could be scanned against
38:11
the database of all images that have ever
38:13
been published, which is, you know,
38:15
competition was still challenging. Something
38:18
that I expect to be solvable so that people
38:20
who want to reuse an image from an outer
38:22
group from an older paper will
38:24
be caught. And that is something I could
38:27
never do. I can only compare a couple
38:29
of images to each other, but. I
38:31
cannot remember enough of them to remember
38:34
an image I've seen three years ago. I would not
38:36
remember that. The image. Sure. So
38:38
yeah, it sounds like the software will play a role, but it will never
38:41
be able to totally replace that human factor.
38:43
I think it's fair to say that you've had like an
38:45
undeniable impact on
38:47
science. But as you look out into the future,
38:49
what is sort of your longterm aspiration
38:52
impact that you want to have? You know, what's the north star
38:55
of what you're trying to work towards. I
38:57
hope there will be more well punishments
39:00
is always a big board, but like, like some way
39:02
that people who are caught doing
39:05
fraud, that there will be
39:07
repercussions for them because
39:09
I feel there's, there's too many cases. I've
39:11
reported to journals in institutions
39:13
where our data was just simply no reply.
39:15
So about 60% of the papers I've
39:18
reported in the past years have
39:20
not been acted upon. These are
39:22
papers with very flawed images. Some
39:25
of them just simple errors that could be addressed
39:27
with a correction. Some
39:29
of them like re outrageous
39:32
Photoshop jobs that are so
39:34
clear to me in five seconds, that
39:36
there is something that is very fishy going
39:38
on there. But five years down
39:40
the road, these papers are still out there. And
39:43
so. I'm looking forward
39:45
to work together more with journals,
39:47
with institutions, with publishers to
39:50
very quickly address these, these
39:52
problems and not have them look
39:54
the other way. There are so many conflicts of interests
39:57
where journals do not want to respond
39:59
because they might lose their citations. They
40:01
might lose their image. As
40:03
a, as a, you know, a good journal that would never
40:06
publish any fraud and institutions
40:08
also do not seem to want to address
40:10
these cases because maybe they have a very
40:12
famous professor who is
40:14
being accused of something bad, but
40:16
he or she brings in lots of money. And so
40:19
let's just pretend this didn't happen.
40:21
And so I'm looking forward to. A
40:24
time where these cases are swiftly addressed
40:26
and where there's much more room to
40:29
give money to honor scientists and not the scientists
40:31
who cheat and that's still a long time. Cool.
40:35
So yeah, we, we have a bunch
40:37
of admiration for your work. I'm sure all of our listeners
40:39
will too. I just want to wrap things up now
40:41
by asking you, what is your favorite. Oh,
40:45
my gosh. One of my favorite papers is
40:47
that of Lawrence David in
40:49
which he. Sampled
40:52
himself sampled his microbiome. So
40:54
took his own stool samples and
40:56
followed himself and another scientist
40:59
for a, about a year. I
41:01
looked at how his, the composition
41:03
of his bacteria in his stool changed
41:05
over time. When, when for example,
41:08
he went camping or he went like people sick
41:10
or went to another country and you can
41:12
see it. You can see the stability
41:14
of the human microbiome. And you can see
41:17
also the periods
41:19
where the microbiome just changes
41:21
because he got sick or, you know, the little
41:23
things we go through over time. And I felt that
41:26
paper was
41:28
so important to show the enormous
41:30
stability of our microbiome. And which
41:32
is amazing because we eat different things every day.
41:34
And so we feed our microbes different, different
41:37
foods every day, but it's pretty
41:39
resilient to the changes that
41:41
we, we we bring along to it. But
41:44
when we have a big change, when
41:46
we got really sick or we go to a different country,
41:48
this is where the microbiome of the human changes.
41:50
And I thought it was so elegantly done. That
41:52
was one of my favorites. That's
41:54
I would love to read that it, could you send us a link and you
41:56
can put it there? Of course. Yes. It's it's an all
41:59
okay. For, I haven't really kept up to
42:01
date with microbiome papers, but so it's
42:03
probably all my questions around eight years old
42:05
or so, but yeah, it's a, it's I'd love the paper.
42:07
Just has some really cool graphs. Oh,
42:09
nice. Yeah. And it sounds like such an interesting story
42:11
that he's telling you to know. It's not just one piece
42:13
of research, but this, this guy's life
42:16
that is being explored through science. I think it's a
42:18
very poetic yes. All
42:21
right. Well, thank you so much for coming on the podcast. Elizabeth
42:23
has been such a cool episode. One of my favorites so
42:25
far, and I've learned a ton. Well,
42:28
thank you for us. Thank you. For sense. Was my,
42:30
my pleasure being here. Thank you. Thank you
42:33
so much, Elizabeth. I really appreciate it.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More