Subscribe
Share
Share
Embed
A discussion with a leading multiomics researcher, Dr. Karsten Suhre of Weill Cornell Medicine in Qatar. Plenty of discussion around intermediate phenotypes, metabolomics, genomics and proteomics, and the increasing role of population health projects.
Proteomics in Proximity discusses the intersection of proteomics with genomics for drug target discovery, the application of proteomics to reveal disease biomarkers, and current trends in using proteomics to unlock biological mechanisms. Co-hosted by Olink's Dale Yuzuki, Cindy Lawley and Sarantis Chlamydas.
Welcome to the
Proteomics in Proximity podcast, where
your cohosts from OLink
Proteomics talk about the intersection of
Proteomics with genomics for drug target
discovery, the application of Proteomics to
reveal disease biomarkers, and current
trends in using Proteomics to unlock,
biological
mechanisms. Here we have your host. Hey,
there. Welcome to Proteomics in Proximity.
I'm your host, Cindy Lawley, with my co
host, Dale Yuzuki.
Today we're talking
to Karsten Suhre from Weill
Cornell Medicine. He's, uh, calling in
from Doha, Qatar, where he's been
driving collaborations for over ten
years. He's here he's going to talk
to us about integrating genomics data
with various intermediate phenotypes
like Metabolomics and
Proteomics. He has recent publications
oh he's got so many publications, it's
ridiculous. Uh, he does meta
analyses on COVID, uh,
obesity, translation of
biomarkers to the clinic. And if we
go far enough back, he's got some
publications on atmospheric science, which I
am just fascinated by, coming from an
oceanography background.
Uh, so this man who
exists in boundary conditions
between sciences and really
overcomes barriers for analyzing
these data is
so exciting to bring
on board. So, Karsten,
with that introduction,
would you like to just start off by telling
us what you'd like our audience to know
about you going into our
discussion?
All right, thank you
very much, Cindy and Dale, for
having me. Um, I
mean, you already covered so
much about what has been going on,
so I would just say let's
just go into it right away.
Okay,
go ahead. I understand that you
came from a background in physics,
like atmospheric physics,
theoretical quantum physics.
What caused you to go into
the genomics field?
I mean, it was a chance event.
I initially did quantum field theory,
went to England, learned about fluid
dynamics, like that, very much, tried to do
a PhD there. But then I moved basically
into atmospheric chemistry.
And then at some
point, I went back to industry
for private reasons and
had to go back to working with
engineers and stuff like that. And then
at some point, I thought it's much more
interesting to be a researcher rather than
an engineer. I wasn't sitting into
engineering. And at that point,
just by a chance event, I read
The Le Monde where they talked about the human
genome, and I just had the chance
that the CNRS actually allowed me
to change field when I
returned to... "CNRS"...
What do you want to say? The
National Research, um,
center in France. And they do
basically everything.
In 2001, when
the draft was announced, that kind of thing
was what you saw?
Exactly.
And I was supposed to go back to
atmospheric chemistry, but I thought,
this is interesting. And then I came
across, uh, a lab in
Marseille, where basically they were all
physicists,
astronomers.
Bioinformatics at the
moment didn't exist, and
bioinformaticians were basically physicists.
So he was learning the
language, he was learning French. He
already knows German. He
knows English clearly, and he's learning the
language of biology. I'm saying it's not
that easy. Well,
also back in 2001, right, the
human genome had just been finished in
a draft form. This is when GeneBank was
still in a library of 15
CDs, compact discs, right?
And this was at a time when,
uh, people's comprehension of
genes was growing,
but we had no idea how many
genes there were, right? Estimates
for 100,000 genes,
et cetera, et cetera, et cetera. But to have
a draft and all of a sudden, okay, now what,
right? What
was that like? Well, at the time,
we worked on a thousand genes. We
do bacterial genomes at the time.
You know, one
genome in science, a single genome.
Yeah. Right. And as far
as that time goes,
what did you find about that transition
from atmospheric chemistry? Help
me. Is that like ozone layer type
of work, or is that another kind?
Yeah, we actually
have a Nature of paper on ozone in the
upper
measurement on
aircraft. It doesn't even
show up in PubMed, because that is
medical. That's funny.
Um, but in the
end, after
that, actually, I didn't really change
fields. The thing is, if you go for
atmospheric chemistry, what we did is
we measured, uh, all kinds
of chemicals in the atmosphere using
aircraft and using ships and
balloons and LIDAR and
things like that. And then when you
integrated the data and that was my job,
basically, integrating all these
measurements. I like to say
it's the same thing as what we do before.
The only difference is the organism is the
Earth. You have
atmospheric chemistry. It's epidemiology,
right, dealing with big data, messy
data, figuring out
what are your cutoffs, what are your
outliers, and, uh, what to believe.
And the systems biology. I mean,
the kind of modeling of the data.
We had, um,
differential equation systems that
are the same, that are used today to model
metabolomics in cells. It was just the
metabolomics in the atmosphere and things
were transported around by
this is great. The systems
biology of the Earth
versus the systems
biology of the human.
Yeah. And you would be surprised that
there's so much match from one to the other.
You could match all the
chemical reactions and stuff like that.
There's so many problems that are exactly
the same. It's it's the
human atmosphere of the human.
I love it. Yeah. It makes
no sense. And
therefore, then is it
CNRS? Then you got involved in the
genomics space. Can you tell me what it was
like in 2002 to
that was HapMap, right? Uh,
Cindy and I were at... No.
Not even HapMap. I
mean, when I started, it was a lab
in Marseilles, and they were working on
bacterial genomes. So they were some of
the first off sequencing them and
actually integrating them with, um,
structural biology. So whenever there was a
gene discovered, the team
crystallized the protein overexpressed it,
of course, did the structure and then try to
understand the function. So it was
functional bionformatics. The
human stuff only came in 2006 when
I moved from Marseille to
Munich. Then suddenly I moved from
bacteria to humans
with the KORA studies (Cooperative Health Research in the Region Augsburg). Yeah, that KORA
study was an amazing study.
You and I were in Illumina, uh, by
then.
Help me with the KORA study. I'm not
familiar with that.
Yeah, well, the KORA study
is a population study. So
basically, you know, maybe the
Framingham study
and KORA is the German Framingham study,
if you want, like, thousands of people
recruited every phenotype, blood
drawn, re-recruiting,
things like that. And the KORA study at
the time was really just the right
moment because all these
onics started off so they were
genotyping them. They were starting the
GWAS in 2006. The first GWAS came
out. Uh, and
then where we were is
like we were the first to actually run a
GWAS with Metabolomics
at the time, with Biocrates and
Metabolome, the two companies.
And from there,
actually, if you go further on,
came the idea to do GWAS on all these
kind of, intermediate phenotypes, the
different metabolized proteins,
you have it. And that was really
very interesting thing. And everything I did
since then is basically a, uh, scaling up
of what we did. And that
particular association... Right?
The GWAS studies that we're
familiar with are GWAS to
disease. You're bringing in
a new layer, and you're calling
it an intermediate phenotype.
Can you explain that a little bit more?
Yes. What is an intermediate
phenotype?
Uh, okay, I should pay credit to
the person who really brought up the idea
that's Floyan Korndirk, he was also
formerly Munich, then Austria.
Um, so the idea of the intermediate
phenotype is to say you have the genome that
actually produces your
effects, your protein levels, your
metabolic level, and then that affects the
disease. So if you want to know how
a genetic variant goes
and determines the disease endpoint,
you can say, okay, this gene has a
risk for diabetes or whatever, but you don't
know really why and what's the function of
the gene. The moment you put
a phenotype in between, that is basically
an outcome of your genetic variation and at
the same time a cause
for the downstream effect on the
disease. You're much closer
to a phenotype, and you have a much stronger
statistical signal once you do your
association studies. And then, of course,
you can build up networks out of that,
right? You can go from transcriptomics to
proteomics to metabolomics.
You could put in protein
glycosylation phosphorylation
everything you can measure on an omics
level, and then in the end, link it
to the endpoint. And I think
nowadays, uh, the concept
of Mendelian Randomization very much
gets into that as well. To say you have this
intermediate phenotype is basically like
what people think as exposures, and then you
have the outcome. Um,
I want to just say
that when Karsten and I spoke
when I was doing metabolomics,
him, uh, telling me that this
story was very
impactful, just like, blew my
mind. That, uh, an
intermediate phenotype in power,
like, I hadn't thought about the power of
it, but it's a bit of a magnifying
glass to be able to help you see
with by improving power,
help you see what's uh, really going
on in the connection between
disease and genetics.
So I just wanted to
honor, um, that
explanation because I think you do a great
job of translating
information into
a biologist's
perspective.
But in the end, it's all statistics, right?
And when you use the word power, it's really
well defined. And
it's the way to say,
if I do a study in a human
population, I have a lot of noise. I mean,
it's not monogenic mice. So getting
the signal out of the noise is
what is your statistical power.
So what chances do you have to get
a statistically significant
signal? And an
association with a complex endpoint
like diabetes needs much
larger numbers than an association with
a metabolic intermediate
phenotype. So you can go stepwise
from the gene
to the metabolite and then associating the
metabolite with the end point. And as
I said before, nowadays you can put in
Mendelian Randomization and even get
asked, uh, questions like is it causal or is
it worth being targeted by a drug? Or is it
rather biomarker? In 2005, right,
I'm selling the first human
microarrays, whole genome microarrays to the
National Institutes here in Bethesda, and
then after the Solexa acquisition, started
selling, right, Genome Analyzers
for looking at all this variation. And don't
skip over the, uh, work we did
with the NHBLI in order
to do the Framingham cohort, which
was certainly one of
I remember staying up till four in the
morning writing that grant Dale!
Well, I was going to say
that all that genomic work
of whatever from
Analyzer coming out, um, to
the present, so much genomic
data, first
thousands, then hundreds of thousands,
now millions of genomes and
exomes, and yet so little
impact upon such common disease. Like
you mentioned, diabetes, right, where there
isn't a genetic signal. But what you're
saying now is genetic variation
with the intermediate phenotype, we can
understand how people get
diabetes, is that correct?
Yes, that's the point.
And also, you can basically
dissect the pathways. So if you have
now 100 different genetic
signals on diabetes,
they don't all associate
in the same way with the same metabolites.
Some may go through more
brain pathway, an
endocrine pathway, whatever. And the
metabolites that are associated with these
variants, they can tell you which pathway
things are going and what's really
happening. So in the end, I also like,
there's an image, this actually comes from
Biocrates. They had this first on their
slides to say in the end, this
intermediate phenotype is like an
imaging. So they had this kind of image
of a dog that was like in very broad pixel,
and the deeper you go into, there more
resolution, the more you see. So in the end,
I think intermediate phenotype is
like an imaging of your
biochemistry and what happens in your
body. And that's why it's so
worth it to go there. Therefore, you're
getting very unusual connections to
diabetes. Then, like you mentioned in terms
of. Well a systems biology piece, and you have
an opportunity, I think, also with a,
uh, non northern European,
uh, population to
characterize biomarkers that will then be
moved to the clinic in order to
monitor diabetes, where
triglycerides or Hb1Ac aren't
successful. Right. That's the
exciting aspect to me is
being able to identify
new ways to monitor
or to predict those that are at a
higher risk for developing
diabetes, uh, years in advance.
But now you're talking about
the Qatar biobank, right? When you say non
European populations and things like that, I
mean, that's a very important point
as well, I think something that's nowadays
also linked to these intermediate
phenotypes. And I think there if we come
back to, for instance, what Olink is doing
in the UK Biobank is this
quest especially of interest.
That is to say we want to find human knockouts
like PCSK9. So we would
like to find a human homozygote
who's hopefully living well,
and then try to figure out why
is that, what does this knock
out do? And is it something that can be
targeted, that can be beneficial, and things
like that. And we just did an analysis on
the Qatar biobank, and it turns out because
we have this kind of very,
um, um,
homozygous population,
basically you have 160 times
chance of finding a human
knockout in the Qatar population
compared to UK Biobank. And that of
course, makes it even more interesting to
go for people where you have
a rare homozygote associated with
extreme protein levels,
low or high or extreme low or high
metabolite levels and things like
that. And then from there on, it's not GWAS
anymore, right? It's basically like these
individual cases. It's not even rare
variant. It's really like you want to have
the homozgotes that are there and see uh,
PCSK9 is this very low
LDL? And that sounds like a very low
cost study. Once you've been able to
identify those
knockouts, you can do it in a small number
of samples, but you have to have the genetic
data. Is that fair to say?
Or does it take a lot of
samples to test
that? I've just been
thinking about GWAS. Right, GWAS.
There was so much we couldn't see
in those first GWAS studies
back in 2005, 2006,
that were done with the GWAS arrays.
I mean, you still need big
numbers. To find it in the
genetic data, sure. But once you've
got those genetic data, I guess my question
is structure
of the
study when you're layering on
proteomics,
could be a smaller study,
because,
uh, you've identified
the heterogeneity and characterized your
extreme phenotypes
or what you expect to be your extreme
phenotypes.
You're
right, you could do that. But I, uh,
would probably still go for the whole cohort
in terms of metabolomics,
because in the end, you don't know who you want
...you don't want to start. I love that
messaging. I love
that.
Karsten, would you mind backing up a bit and
tell me a little bit of higher level view of
this particular study or this
particular cohort in Qatar that
you're working with?
I just said
KORA is the
Framingham biobank.
And I do want to say
that Peters is the current
PI of KORA cohort
there out of Augsburg, right?
Uh, she's in Helmholtz.
I just want to give her a shout out.
But yeah, back to the Qatar
biobank. Yeah. So how
large is it and how long has it been,
uh, collecting samples?
I think when I came here, they
were building it up in collaboration
with, uh, Imperial College. And
actually there were also colleagues from
Munich who were coming over here as
consultants. So they get in contact with
them pretty easy. I, um, think at the
moment, they are well above
will fully sequenced.
Uh, we ran
3000 samples on an
in house metabolic platform here
in Qatar. Also 3000 on a SomaLogic platform.
And we started working on that. And I
think there's a huge potential there
to potentially do the same thing
as what is happening with the UK
Biobank finding. Maybe someone who's ready
to put in funding to get more of its
data and things like that. That's
pretty exciting. And as I said, this
specific, genetic specificty, of
course, makes it very valuable
to discover these human knockouts. To be
clear, then it's agnostic for
disease in terms of
you're looking at 20,000
individuals to the metabolome,
the proteome. Like
Framingham. It's just a cross
sectional thing. Of course,
Qatar has a very, uh, high diabetes
rate, which is a challenge for
Qatar. That's why Cornell is so engaged here
in Qatar. But it's also, of course, a chance
if you want to study diabetes, because you
have a very high there's over
have a lot of undiagnosed
diabetics in there, which is also
scientifically very interesting, because
these are people who are not treated with
medication. So if you do all the omics on
them, the problem if you want to study
diabetes with omics,
especially in KORA, sometimes
the people with diabetes,
they are much more healthy than
the others. Because they know they have
diabetes, they. Do everything
they need. And then it's
very hard to actually. Study anything
about that a motivation
component, because they've got this sort
of this
deadline by which they need to
reverse whatever was going on before. That's
really interesting. Yeah.
Now I remember, there
were leaflets from diabetes associations in
Germany. They actually said, don't worry,
diabetes is not necessarily something
negative. I mean, of course it's a
disease, but if you take it the right
way, it could actually change your
life
plan. Right. And prolong your life
That's fascinating.
I mean, I wouldn't
go and get diabetes. Does
speak to the power it
speaks to the power of feedback. Right.
And some of these companies that are trying
to provide a, uh, way
for you to monitor your health, that's
not going to be paid for by insurance, at
least not in the US. But
it speaks to if you
identify something really early that looks
like you're nudging you're out of
normal or healthy
range, then there's a motivation
factor to nudge it back. Right. It's a
wearables argument, maybe.
Especially in this diabetes field. Of
course, the problem is if you're actually
able to change things, because there are
some things not everybody
can change like this. And there's a lot of
brain component in this as well. It's not
just like, oh, you're fat, you
lose weight. It's
much more complex than that. And it's
interesting, the diabetes to heart
attack or heart health
connection, because I was just reading a
paper last week that talked about
a particular drug on
market for diabetes that
reduced the risk of heart failure. And
it's just remarkable. Right. In terms
of we get back to system
biology. We have
a connection. Systems.
Yeah.
As far
as this particular study
goes, that you're
working with the Metabolome and the
Proteome. Uh, is it you're
now looking for funding or
you said that
it's like the UK biobank and that it
could lead to something like that.
Yeah, I mean, it's not me.
I don't
think it's from Qatar Foundation. So
actually, I try to
motivate people to say, do
this kind of a deal. It would be for
everyone. Um, but
it wouldn't be up to me, I
think. Moving on a little bit,
I think I'm aware of this
paper that was recently published in
Frontiers and Immunology. What, uh, can
you tell me about it? This is
identification of robust protein
association with COVID-19 disease
based on five clinical
studies.
Yeah,
That is the outcome of...
COVID has changed a lot of stuff
also in Qatar, and it
made us collaborate strongly
with New York, because, I mean, you
know, especially wake up. And in New
York, they are affiliated with the New
Presbyterian Hospital, and they were totally
on the front line at the time, and they
collected samples from COVID
We collected samples here in Qatar. And
my colleague Frank Schmidt, who's running
the Proteomics core here, and he has an
Olink platform at the moment, um,
he brought basically
in these samples, and we measured for
Cornell in Qatar and New
York, these samples,
and then asked, uh, this question
I think everybody was asking at the time,
what is special about COVID? What can we
learn by doing things? And I mean,
there was a big rush, and I think there's
also a lot of not so good papers out there,
so you have to be careful of
what's there. They're not saying ours is one
of the best. It's just like, uh, to
combine the data. And at the time,
I think Olink and you can maybe help me on
that they shared of, uh,
their 1500,
publicly available, downloadable and
MGH study with, uh, several
collaborators. Yeah, I
see 306 cases,
General on the
Olink Explore 1536. So it was a
very broad study.
And then the other four studies that you
mentioned from Imperial College of London,
and like you mentioned, New York
Presbyterian, what, maybe 50
to 200 cases or
so, but, um, nonetheless,
quite, uh, an interesting data set.
Yes, indeed. You're going to say something?
Yeah,
that was
the motivation to say for us, it was an
opportunity to really learn, also learn
about the new Olink platform with the
together. And the idea was to say,
okay, especially with COVID, the
conditions are so uncontrolled because
the situation was just like, oh, grab every
sample you have there. But you cannot make
sure that everybody is under the same
conditions to say if you combine
five different studies to say what comes
out and what's really in all studies at
the same time. Which doesn't mean that the
others, uh, are not relevant hits, but these
are the ones that are really so robust that
even under worst conditions, you
can see them. And that makes them very
interesting, of course, as markers
pathways for
the research. We had some that
were plasma, some were plasma
EDTA, some were inactivated
with heat, others were inactivated with
triton X, a detergent. Some weren't
inactivated at all.
And I'm like so you combined
all these different centers, different
storage conditions. I, uh, don't have
information on how they were stored or how
quickly they're at room temperature or what
have you. What, uh, can you comment on
that from the reviewers
point of view? Right. Don't they raise
eyebrows? What are you trying to do here?
There's too many variables involved.
Yeah,
right. And I mean, it's the same thing in
many papers, that there's always concern
about study limitations and things
like that. But I think as long as you
acknowledge these limitations and you
say, and you go for the really strong
associations and of course, you must make
sure that you don't create bias
by having confounders,
that's the risk. You have to make sure that
there's no confounding. But once you're sure
about all that, then
having something that sticks out of this
noise is really a strong signal. And it's
really worth following up because you're
sure it will be replicated in the
next study and not it will disappear.
Because the next study is maybe not
clean, as clean as the first one.
Yeah. The Abstract this paper
indicates 13
proteins were significantly
associated with COVID infection
compared to controls. And
that these differential expression
of these 13 proteins
was across all five,
which I think is pretty
like you mentioned, it's a strong signal.
And it was
Bonferroni corrected, so corrected for
multiple testing as well. So it's not like
that. Why don't you
comment on this, that important
point, which is the Bonferroni
correction and how many signals
disappear
and that frustration.
But it needs to be
done, right?
Well,
it needs to be done. I think
the point with P-values is always,
I mean, if you have
a really good study that has enough and the
word power comes back, then
you can allow yourself to really go
for the strongest hits
and focus on them first.
If uh, you don't have that, then you have
to do P-value gymnastics, things like
that. Although you should be aware, of
course, I mean, there's a lot of more signal
below that you don't want to lose, right? I
mean, later on, once you know that your
study is good and that what other
caveats... you can look at the hits that are
not that strong and see are they
biologically relevant, can I take them maybe
as a starting point, as preliminary data for
the next brand, and
then reinforce that and confirm it?
So I'm not saying you should only look at
Bonferroni significance, but coming from
the GWAS field, I mean, it's even worse.
You have to be Bonferroni, and
replicate, which is the decision would
say, uh, it's evidence should
replicate, but even then, it not always does
replicate. And that means
it's always good to be very conservative on
these P-values if you
don't want to end up with things that, uh,
just don't work and don't
replicate, because that's frustrating.
If then you follow up on something and then
things don't replicate and you just turn
data up and down, it just refuses to
replicate. One of
the interesting things about the
results, right, in terms of these 13
proteins, maybe some familiar
suspects like IL6 and
Interferon-gamma, uh, what can
you comment in terms of then, the practical
usefulness of these inflammation related
proteins and its association with COVID?
You put me on the spot, because that's
really a medical thing. Of
course, I try to figure out a bit what's
interesting about these. Um, you may also
be a bit more reductive and
say, well, maybe there's a lot of stuff you
would not even be surprised about to find
them. Is it? And I think that's an important
thing. We have other papers where we compare
not COVID case control, but
COVID against bacterial ARDS.
And you ask yourself what is really specific
to COVID Because just to say there's a
cytoskines storm and that's something that
could also happen if you have a bacteria
infection. Um, I
think what really changed something, I
think, about COVID in the last years is
what we learned about how to treat
COVID. I'm not a specialist in interpreting
this, and I find it sometimes interesting
what people actually see in our papers,
which I didn't even see. And it
would help doctors, I think, to
understand what kind of
excess you really want to want to go in your
treatment. I think my understanding is that
a lot of why the death rate in
COVID went down over the months and
the years, in part, is because people
understood better how to actually
treat it. And this kind of information, I
hope, contributed to
this kind of understanding, at least.
that's really helpful because, uh, to be
clear for this audience, ARDS is Acute
Respiratory Distress Syndrome. Right.
And this is where you mentioned
bacterial infection
will bring about this. A COVID
will also bring about this.
It's basically like a control
because you always want to control, but a
non-COVID person is not a good control for a
COVID person, if you want to know.
So then the person who
has the same symptoms but not from
COVID that. Would be a much
better, like the
Michael Philbin study as well.
Right. These were patients that came
into the hospital with symptoms
that seemed COVID-like but tested
negative. Their controls were just
that, which I thought was
worked well with what you were
doing.
Now, as far as then, um,
the proteomic analysis goes
with these particular
signals, is there a particular
translational message in terms of
applying sort of this
knowledge clinically, or
is it still too early? When
people have severe COVID,
it's already known, it's
diagnosed, they're monitoring for this. Like
you mentioned, it suggests new treatment
ideas, new clinical
trials.
I was just going to add on
to that question and who might
pick this up from
those that you work with? Who might
pick this up and take it over that line?
Right.
Uh is it
Gabby,
just
not to put you on the clinical spot?
Yeah.
No, I think the
point is, in this case, we would probably
not even know because the people who really
pick it up would be the doctors who are
capable of reading these papers and
translating that into
practice. And maybe I don't
know whether it may be too
dangerous, but maybe let me
comment carefully on this. You may have
heard about Didi Raoult. Can
you he's the
person. Give us some background. Yeah,
he's a
French doctor who
promoted. The use of
Ivermectin. I'm
not. Yes, actually,
careful. I don't want to make any
statements here. I've been working directly
with him. We worked together when I was in
France and we sequenced bacteria and stuff
like that. I think the way
when he said, this works in my hands,
he was criticized a bit against
whether his studies really worked or not.
But knowing him, I think he was a bit like
he treated patients
in a way that he knew how to work with
these patients. I think the knowledge he,
as a doctor personally had,
and that may come out of, uh, which,
uh, chemokines and
cytokines and whatever work, I
think in a way, he may have
treated patients right the right way and
treated them. What didn't work out was
this statistical thing like black and white,
broken or not. And
that's a whole different story. But I
think my feeling he was
very honest on that end, that, uh, he
probably for himself treated this patient
and saw an effect on that. And
now coming to this translation thing, I
think actually based on this kind of
knowledge. So I think interacting with some
doctors is like, there's a lot of
doctors don't really work like a
robot. Like, say, if there's A and do B
and then do C, there's a lot of more
like, what's the whole feature of this
patient? And that's where the omics
comes in. And I think with the bridge, it's
still not there. How do you really bridge
that in a way
you cannot necessarily
nowadays, easily bridge? Maybe
artificial intelligence at some point,
algorithms, things like that could do it
algorithmically. But I think there's a lot
of knowledge, stomach
knowledge from doctors, where they
learn from these studies. They take
something from there. They have their own
picture on what's going on,
and then they treat
patients on their basis. And it's
very hard to quantify and translate it into
a case control research paper.
And that's where maybe where Didi Raoult
get into some trouble
because people expected that to be a
case control study.
Is it the role
of the MD, PhDs,
perhaps, to do that, or translational
scientist is, uh, a job description I see
much more commonly now. Right.
And I'm trying to
imagine what are their
objectives and
what are the key results that they're
evaluated on. And I hope that it's
helping facilitate this
translation, because I think that when
a scientist throws it over the
fence, it,
uh, may not land in a place
where a person can
carry it and then to the clinic.
And I think there's a
huge gap between what we do on the research
side and really getting it to
the doctors. Because the doctors, in the
end, they. Want to have something that's
replicated, telling. Them, here are
It's not something they can really
translate. And
that's really the big thing where I think
we've been working a lot on it. I hope
things are getting better.
But if you want to go to personalized
medicine, you cannot go at
the same time, to double blinded clinical
trials. But it is an n equals
one thing, if you want to treat one
person, you cannot do a test on
a hundred before you decide
on the doctors. Have to make those
decisions, which always, uh,
seem challenging to me.
Yeah, one of the interesting points you
bring up is the art
of medicine versus the science of
medicine. And like you mentioned, in terms
of treating the individual patient through
personalized medicine, oftentimes go
with gut decisions based
upon decades of experience in
treating patients with similar
symptoms. And then it's,
well, okay, what is that
art? What is it about just
patients? How do you capture that?
I think
that's where, of course, all this omnic
stuff, if doctors learn about that, if they
can integrate that into their gut, uh,
feeling in a certain way, and then, of
course, this gut feeling and then
with the other buzzword, if this gut
feeling could be translated through
artificial intelligence or whatever kind
of things into something
there. Because in the end, you don't want to
have, like, two doctors in the world who can
cure everybody, and the others don't know
what to do. Right? You want to
basically have every doctor be able to cure
every patient. And uh, going
back a bit to a topic we touched on
a little bit before in terms of that
intermediate phenotype multi-omics
kind of approach, it seems like your work
now primarily focuses on
proteomics and metabolomics.
Uh, you mentioned, well,
you can go into glycomics, right?
You could go into
transcriptomics, methylation or
epigenetic methylation.
Yeah. I'm curious, particularly on the
transcript piece, your
perspective now in terms of the usefulness
of RNAseq, either at single cell or bulk
level as it intersects with this
intermediate phenotype idea. Because
obviously DNA to RNA to
protein, is it very
useful signal in terms of RNA, or is
it sort of one of these depends on what
problems you want to answer. I'm interested
in your perspective on it. I'm single-cell
versus bulk. Yeah, it's fascinating.
Oh, yeah. Uh,
first of
all, transcriptomics,
have to be careful about when you do
metabolic cell protomics, we look at what's
separate in the blood. When you do
transcriptomics in the blood,
you look at the transcriptome also by
blood cells.
So that already doesn't link to each
other. Um, I
think the other part where
transcriptomics comes in, and a lot of that
is GTEx, where you're basically
transcriptomics of the organs. And that data
is hugely important and
it always gets overlaid with GWAS data
and things like that.
But I'm very much a big fan of
transcriptomics and white blood cells. I
think there are not many people really doing
that. I mean, there have been a few studies
that data is out there in GTEx. It's used,
um, but further than that, I mean,
it would be really like, uh, and
it's very much confounded with the white
blood cells and things like that.
Um, we've done work with methylation,
so methylation is
a good proxy for transcriptomics in a
certain way, but it's also in the white
blood cells if you don't have access to
tissue. And then Cindy
mentioned single cell.
Now single cell is a total
different beast, right? I mean, single cell is not
something you do on the population level
in KORA. It's mechanistic
biology.
Uh, I think
single cell is basically once
you get ideas from
GWAS on potential pathways, things like
that, then it's the next level
where you try to really
pinpoint what's going on, what's going on
in these cells and things
like that. So in the end, I think
there's like terms that really need to
be put into context. And the
same is true for metabolomics and
proteomics. Of course, there's metabolomics
of cell culture, there's proteomics of
tissue extract, and we
haven't even mentioned the microbiome. Exactly.
But also, what do you think about it? We've
just got a couple of papers that I can think
of that have done this, but running,
um, say, proteomics or
metabolomics at the same time that
you're taking, um, the
single cell measurements.
The single cell transcriptomics
measurements. Or CITE-Seq, which also
includes the proteomics of the cell
to be able to identify cell type
and being able to see over time
what's showing up in the plasma. That seems
really compelling to me just in terms of
translating something mechanistic
into something you might be able to measure
in the plasma. But
it's all discovery. Right?
Well, I think probably
the first nearest thing would be single cell
proteomics. I don't know where exactly that
stands, but I think doing bulk proteomics
and combine it with single cell RNA and you
have it. That is not really
something I favor. There are other
techniques where you can have like, um, you
can have this kind of antibodies
for this protein surface and have single
cell resolution like that,
as data like that. That of course
makes sense. But
also in terminology, is that really
proteomics? I mean, you're doing
antibodies or surface proteins. So even the
term proteomics, you have to be careful.
Yeah. Single cell transcriptomics where
you're able to identify a cell type.
That's how I think of it. Uh, although
people do call it proteogenomics.
No, there's also protein things you can.
Do from the cell surface. And not
only is transcript
yeah. So then you
can determine if it CD4 positive, CD4
negative, what's the nature
of those cells. Right.
And then I think being able to
say, uh, the publication that comes
to mind is the
Italiano paper on LIF,
uh, leukemia inhibitory
factor, uh, where they looked
at, uh, PD-1, I believe it was PD-1
an immuno-oncology
treatment and responders and non responders,
and simultaneously did proteomics in the
plasma. And that, I
think, was what I had in mind. And
there's maybe three publications I know of
that had that. And,
uh, I don't know, I guess I'm biased
by trying to figure out what we can measure
without
biopsy, uh,
and being able to make as much use of
those precious biopsy samples as we can to
then translate it. But you also talked about
methylation, and I think I diverted you a
bit.
Uh, no,
I just brought up methylation
because you also did not only with GWAS,
but also if it
was with methylation. And
there are also see coming up where
I think, in my view, we have this
kind of thing like, say, genetic
variants in your metabolism. We like
to say these are like different bias
settings. You have your organisms, like
all these pipes, and your
metabolisms flies through the system, flow
through the system. And then these biases
make you individual. That's your metabolic
individuality.
You don't tune them,
you inherit them.
And then,
um, um,
your metabolic
status, which is based on your genetics
and to some extent, your environment
is tuning them, I guess, was what I
was thinking.
I mean, there are two things. The
one is that comes from your parents. That's
your genes. So that's what I said.
And you can't change
that. Then, of course, the system
buffers itself and regulates itself. And
then you have other regulators that could be
the methylation for, because
then there's a certain protein, you
need it more or less. And then the
methylation adjusts and says, I need more
or less of this or that enzyme. And
that's something we've seen in some of the
studies. And then you have the
methylation. To my view, is a bit the
readout of the
stress and how your body reacts
to disease. And in that sense, it's a very
complementary tool to the genetics. And you
can basically get the nature
versus the nurture in one measure. And
you can do that in population
studies with the
arrays for genetics and
for methylation at the same
time, and then combined it. And we've done
that. We have papers with metabolum
and with metabolomes
and with proteins on them. And
especially for diabetes, all these complex
disorders, they're really a nice signal
coming out. And priority
of positive control is like, smoking
is one of the
strongest methylation, um, signals. And you
see, like a handful, uh, of
genes that are all associated
with smoking. And these genes
make sense. They actually are in the
pathways for (unclear) de-toxification and
things like that. So I think there's a lot
of promise of
combining these things as well, like the
methylome and the proteome.
And you mentioned also the
glycans. So we have glycan data as well.
So you can add protein
glycosylation on that and find out what's
the point IgG
glycosylation, what the information says,
things like that. That's
what fascinates me to get more and more
data, and ideally all in the same
sample, because then I can correlate the one
to the other, and we understand
how they interact, how they work together.
And then the GWAS on all of them. The
GWAS on them because the GWAS in a way, is
also a way of actually
questioning your data. How well is
your phenotype? Because a genetic
variant can lie. It's always causal.
Other factors can be
confounding or whatever. But genetics,
except for some very constructed
cases, it's always going to cause a variety
unidirectional. Right. You don't have to
reverse. Yeah, very good
point. Yeah. Your
genes don't change because the metabolome
changes. And I
admit, I did open up a big
box with that particular
question. I want to thank you, Dr.
Suhre, for your generous time. I have
one final question for you.
And how's the weather in Qatar
today? It's September.
Is it still hot? Yeah,
it's
something around 40 degrees, but it will
go down, and there will be the world
football version. There you go. November.
And then it would be nice, like, 25.
Can I count on you for, uh,
after World Cup report? In terms of what
it was like, I
don't know anybody else I don't know
anybody else in Qatar at the moment.
So you will be my Qatar
representative?
No, I mean, we're expecting
between one and 2 million people coming to
Qatar,
and you can buy tickets at the
moment.
Do you have
tickets yet?
No.
Honestly, we got to add it
to. Say that
I hope you wouldn't no, it's okay.
It's all good.
But Qater has been preparing for ten
years for this, and I think
they really put a lot of
I mean, they're really
passionate about that
in the west. Very often. It's like, oh,
gosh, about the football club. And there's
negative things. Uh, sometimes it's
frustrating to hear my compatriots
in Germany oh, you have to boycott these
guys exploiting...
Absolutely. When you hear
passion, it's true
love. And especially,
I think, the emir, he
really brought the football cup here
because. Football fan,
it's kind and generous of you to leave that
ticket for someone who's truly.
Thank you. Thank, uh, you for the
conversation. I really enjoyed it so
much. Bye. Thanks, Carson. All
right. Bye bye. Bye bye.
Thank you for listening
to the Proteomics in Proximity podcast
brought to you by OLink Proteomics. To
contact the host or for further information,
simply email
info@olink.com.