So Greg joined us about eighteen
months ago from the University of

Queensland in Australia where he'd spent
eighteen months of extended sabbatical.

Having spent ten years before that at
North Carolina State University where he

stabbed himself as a quantitative
genomics of mainly working under software

last few years he switched to studies of
human genetics and he came to Georgia Tech

to establish a program and a Center for
Integrative genomics here a program built

around the idea that systems biology can
be used to sort of bridge a gap between

every linear epidemiological things at the
population level in the sort of cell of

molecular biology and
engineering that we do here Georgia Tech.

And today he's going to
talk about that interface.

Thanks so much Bob So Pleasure to be here.

So as Bob mentioned so
I work on religion I mean epidemiology.

These days.

So what I mean is that in is is is
the consequences of all different

environmental influences and

how they interact with the genome when
fluence phenotypic and disease stepped in.

If it diversity in
disease susceptibility So

when people move from rural areas
to small towns order big cities.

We know that that actually impacts disease
susceptibility in many ways whether

schizophrenia or metabolic disease and
we know that events

that happen during a person's life such
as economic crashes impact their health.

We know the parental effects and

increasingly we see that grand parental
effects have an influence on health.

And of course we know that there's

an enormous component of
genetic susceptibility.

And in the last few years there's been
this incredible revolution in human

genetics where we have the capacity now to
probe the entire genome and pretty soon

we have the capacity to sequence entire
genome three thousand dollars or so of.

To really get at the genetic basis
of variation but of course if you

as you're all probably aware the findings
of from that are a mixed bag because

we have this microscope that gets
in now to individual variants and

there have been thousands of variants
found for hundreds of traits and

it's an astonishing advance but
the cup half empty view

is that it suddenly explaining ten or
fifteen percent of the genetic variants.

Even with samples of
fifty thousand people.

OK So there's a lot missing.

But the other thing that's
missing is the environment.

OK so we know that genes
interact with the environment and

my perception is that if we're going to
understand that we need more functional

sort of our size that sort of don't just
go from genotype to disease risk but

open up that blocked black box and

go from genotype to gene expression
gene expression to protein abundance

protein abundance to metabolite
abundance and so forth.

So that's what we're going to try to do in
the center here is add those components.

OK.

So I wrote a book a couple of years
ago called It Takes a genome.

Built around the idea that there's been
a dramatic shift in our environment

in the last two generations or in fact the
last generation in many parts of the world

and there's been dramatic shifts
in our genetics over the course of

the last hundred thousand years or so
as we've migrated around the world and

the cons as a consequence of that
the human genome is not in equilibrium.

And my conception is that diseases are
rising because of that imbalance between

our genetic heritage and our modern world.

So we're seeing this demographic shift if
you like of disease susceptibility from

being infectious diseases and

diseases about malnutrition to globally
now the disease of the affluent

cancer diabetes asthma depression
cardiovascular disease and so forth and

the real question is what why is that
because our genome genomes haven't changed

that much in the last generation or
so but our risk certainly has.

So what I'll do today.

Is.

Stopped by presenting the first
half the talk will be two

sets of studies we've done that have
got us into the human genetics and

one approach the first study is what I
call geographical genomics of this human

transition in lifestyle and
I'll discuss signature of the effect

of environment lifestyle on our gene
expression profiles in peripheral blood.

And then I'll describe a genome
wide association studies for

gene expression in the blood
that we did in the sample there.

And then a second study.

This is stopped there we go is motivated
by this concept of fifty phenotype

which is the notion that
maternal health is impacting.

The child's health already a birth and so
it used an expression profiling to get at

that and then I'll spend the last
fifteen or twenty minutes talking about

integrative genomics what we're doing at
Georgia Tech and how I see what we're

doing interfacing across side B.B.
and the university as a whole.

OK So you say if you doctor or

here was a graduate student in my lab
who came on a Fulbright fellowship and

I talked him into going back to his home
country Morocco and doing this study and

he took at and ran with it and
did an absolutely tremendous job.

Not just in my view but a lot of he's
won several awards for this work and

it goes like this that we decided we
wanted to sample people who were living

different lifestyles and the bottom line
is the idea that our disease risk is

going to be different if well with
we're living sort of traditional human

lifestyles whether we're desert nomads or
Bedouins or for living in very small rural

communities or if we're actually living
in modern cities in this case in the slum

of this city of agony here on the on
the Atlantic coast of Morocco.

So I like to sort of just distances and
in the morning We're all waking up artist

and like about this Google Earth.

Flyer the video to sort of illustrate
Morocco which if you get a chance.

I recommend going

there because it's one of the most
beautiful countries on the planet.

So we went in and we started in
agony here which is this city here.

So there's a beach which is full of
English tourists getting some burn and

we go about two or three miles north past
a concrete smelter there that you see.

And then you have this.

Dense urban So low income community here
by the ocean with fantastic views but

you know very bad Risperdal three
problems that air pollution

very crowded there through
downtown Agha dear and

then we come to cheer our Which
is another sort of low income.

And dense region next to the airport
here where we sampled and

that the communities in both sides
of the are a mixture of Arabs and

Berbers or form a Maziar people.

So the Arabs came into Morocco in
the Middle Ages so what six or

seven centuries ago and
they typically live in the desert valleys.

So this is another city.

It is neat and it's in fact about fifty
meters from where Yousef grew up so

they live in that the Dry Valleys.

In a typically fairly segregated from
the from the blue truck villages from

the villages who are living higher up into
the hills of the low Atlas Mountains.

So that's a gram where you said
spent his childhood in the summer.

There's no electricity whatsoever
to just get discussing it.

Now running water had to come from
from nearby streams and so forth.

This is the first village we
sample from setting a bore.

It took him six hours on
a donkey to get up there.

So that's fairly remote.

And then the second village is blue truck
that we sample from is also a burger

village one hundred percent Berber there
is a road as you'll see going through

on Google Earth here but
that went in literally.

Two years ago.

So again a very isolated village and the
other place is about twelve hours away.

By four wheel drive way out into
the middle of the Sahara Desert

where we were lucky enough that we were
able to come across about a dozen small

family groups that sort of move every
three weeks move to move their tents so

you saw one in that slot
a couple minutes ago

incredibly friendly people you know
wanted to share a goat with us for lunch.

Even though they only have five or
six minister amazing amazing experience.

So that's Marco.

And actually I'm so excited I was
able to send you know a picture

back to my lab from from you know
an hour out of the desert from right

there who would have thought that I got to
middle of Manhattan you can't get service

on a cell phone at all but
in the middle of the Sahara you can't.

So the question we asked was then how
different is gene expression between

these people and we were sort of thinking
that maybe we'd find a half dozen genes or

a few dozen genes with significant genome
wide and we teamed up with a guy called

John storey who's one of the leading
statisticians working on

analysis normalization of gene expression
data and he used something called

surrogate variable analysis to really do
this properly but it doesn't really matter

no matter how you look at it you see that
there are these different groupings.

So it turns out that the villages the
remote villages are the most different.

So you can see all of those
remote villages here and

blow the clustering together
with a couple of exceptions and

then pretty much the better one living out
in the desert moving around the nomads

leap moving every few months
a few weeks or another group.

There's a subgroup within here which
is in fact some of the women better one

compared to the men and
then there are the people and I get Dia So

there's three different groupings so
you sort of mediately says Your

lifestyle is having a major impact
on your gene expression profiles and

how many genes is it what we do
the analysis of variance on it.

It turns out that in this analysis about
a third of our transcript on is different.

According to where you're growing up.

So we adjust this for

ethnicity and so forth and that's a genome
wide significance levels which means we're

looking at ten of
the minus six ten to minus.

Seven P. values on those genes.

OK that's not just a few
genes it's a very big effect.

And it's also not random.

So if you go to things like ingenuity
pathway analysis and they've they've

annotated the genome with respect to sort
of different DEET disease classes and

other types of gene ontology you see that
the biggest difference is in fact was

genes classified as being involved
in a spiritual disease separating

the urban versus rural samples and
that made perfect sense because

that urban sample is suffering from
a very high level of spiritual illness.

We were wondering at that time whether
this might also engage with epigenetic

effects so we just did a methylation
profiling using at the time

the cancer methylation
panel that alumina had so

this was only about fifteen hundred
genes at that time that had just C.P.G.

islands that you can do the sort of
conversion becomes a genotype in chip.

But it's quantitative and allows you to
estimate levels of methylation in front of

these fifteen hundred genes and our
positive control is in fact a sex effect.

So there were about one hundred fifty
genes that were different between females

and males and they turned out that every
one of those was on the X. chromosome as

you'd expect because you have differential
methylation of one of the X. chromosomes.

But if you look at the last style effect
there were only about two dozen genes

that were different
differential methylated whereas

you know as I said there were several
thousand genes differentially expressed so

it's not obvious that methylation per se.

At least this low level resolution is
responsible for those differences although

I have to say that one of these genes is
in fact the transcription factor one which

is one of the major regular transcription
Regnery factors in the peripheral blood.

And in fact it turns out that targets
of L. Kwan are in rich among genes that

are different expressed so it's possible
that a few changes could actually.

Ramify and effect a large
amount of the transcript and

I doubt that that's the case but
it's certainly possible.

But the bottom line there is adept
enough occasion by methylation.

At least.

Level of resolution is
on the accounting for

a small fraction of the expression
divergence between the US work out of this

doesn't mean that histone modification
micro on a change is and so forth

are also involved but it would take a lot
more work to sort of follow that link.

So.

We decided to follow up with
a much larger sample so

that first study was only a few dozen
people from each of the locations.

So we went back about a year later your
half later and decided to get two hundred

people because it just appeared that you
can do genome wide association studies

gene expression and two hundred for at
least some Lymphoblastic people looked at.

So we gathered about one hundred samples
from from both sides of agony and Shira.

And about fifty from this
Arab village of the gram and

about fifty from this
village of good traffic.

Now there are some I am
a seer living in a gram.

As well and most of the man in the
actually commute into T.'s need to work so

their lifestyle difference is there so
we got prefer blood

samples which we immediately
to Fraction eight.

Leukocytes out so it's we have a structure
nation filter that removes the platelets

and the red blood cells that go straight
into our name later on the filter so

it's preserved we ship it back to to
the states and David Goldstein group

at Duke University ran through
the gene expression profiling for

us on these human H T
twelve arrays which have

forty eight thousand different transcript
probes representing all of the genes.

Well than all the refs genes.

Plus plus a number of others with many of
them with model probes and then we also

did the genome wide genotyping on these
with the six ten quad arrays at that time.

So six hundred thousand which after
quality control boils down to

about five hundred thousand snips
throughout the genomes that we can type.

So the question is how does variation
in Jena types affect variation

in the transcripts given
the ethnic variation and

the geographical variation
that we're seeing.

So when you do this.

I was a study is you have to control for
ethnicity.

So these are what we call principal
components analysis of the genotypes now.

So we're looking at the gene or
to pick structure so what impact.

Does your genetic diversity have on where
you cluster with relative to other people.

So the first component of you need genetic
variation actually separated about

a dozen people now from the from
the other two hundred or so in the study.

So these people and
these three individuals here.

I guess I've been politically incorrect
was a label them as black but

they actually sub-Saharan Africans and

the others were self reporting as
having some sub-Saharan ancestry there.

So we think that that's probably what the
Axis was and we threw in some you Ruben

samples from the hat map project and sure
enough that axis is the sub-Saharan one.

OK So there's just a small component just
a few percent of the diversity but a few

individuals that I would contribute that
the remaining diversity is clearly along

the second axis so if we flip that on its
side here we've got the second axis and

then the third axis of variations down
here and that's clearly separating Arabs

from burgers or atmosphere because so
these are the remote Berber village.

These are the Arabs in
the city of academia.

These are the burgers in
the city of Arcadia and

these the Arabs in the city of Iran.

OK So we're expecting that you
would actually have the the Arab

village one extreme the Burma village and
the other and

then the city would be sort of ad mixed in
the middle that actually can turn south

the probably the Arabs in the city
sort of more pure Arab and

you've had enough admixture with just
a few individuals exchanged per generation

over the last twenty or thirty
generations since he has moved into Rocco

to actually bring the villages
closer to the Berber OK so

you've got to shift over along
the axis and then this axis here that

separating out the different village is
actually largely attributable to the one

portion of the first promise on
sort of not fixed differences just.

Minor earlier frequency differences in
that region but there are enough to give

a minor component a variation that
looks like a separate people.

OK so we can actually take these measures
of phenotypic variation we fit them into

our subsequent models and we control
the population structure when we do our

various statistical models of
gene expression differences.

And I actually worked with statisticians
in the SAS Institute So Russ Wolfinger and

calcium a class on that and then Pete
of issue at University of Queensland in

Australia was sort of sort of
world expert on that methodology.

So the first result is that we replicated
the initial finding perfectly.

So it turned out that the village of.

Boots for the birth of village the very
remote village was totally different from

the others in this case almost half of the
transcript and if you take a false cover

a rate of ten percent is different
compared to the city of academia.

Which is about two hundred
thousand people and that.

Our village of gram.

OK.

So it's a village it's a rural village but
it's close enough to the city.

I guess that you get that last part of it.

Shifting in expression.

We want to know whether lifestyle or
ethnicity was

having a bigger effect on transcription
and it turns out in this case that.

If you just look in agood dear ethnicity
and gender have almost no effect.

So that any of a few dozen genes
that are different by gender.

There's maybe five percent
different by ethnicity

smart a component component
whereas if you look across

the entire southern rock or regional
called the Seuss then you start seeing

you start being out of partition out
ethnicity and location effects so

we think there's probably some combination
of and some contribution of ethnicity but

mainly it's probably geography and
that's confirmed in the next slide.

If we now look at the principal components
of variation for gene expression.

So that one a couple of slides ago was the
principal components of variation for Gene

a type this is now a gene expression of
the people of differing along this axis.

Across all populations and
in fact that access.

We now think differentiates people in
all populations we've looked at in

Brisbane and now in Atlanta we
see variation along this axis.

The second axis is somewhat specific
to the Moroccan population so far.

What's intriguing is you get a very
clear separation of these people

from these people.

And who are they.

Well these are the burglars from
that remote Berber village.

So they're all quite different men and
women.

These are the women from the the Arab
women from that Arab village.

OK So the Arab women and
the remote villages have got sort

of one type of profile and then
the green of the residence Vaga DIA city

the open red circles here
are the men from the Arab village.

So we knew their community.

So that's not necessary
surprising they got that one but

what's really weird is that the women the
burble women from that our village were

classroom with this group as well.

So there are very complex Geographic
by ethnic by gender differences

there are affecting the whole
genome gene expression profiles.

So that's a surprise.

But it just brings home
the fact that expression

profiles throughout the genome
are modulating but by our lifestyle.

Our next question then was.

Are those lifestyle effects robust.

As a genetic effect on gene expression
robust to these last part of it.

So in other words do you see an
interaction between your gene a type and

the environment in the way
they regulate gene expression.

I meant to present this are so
we can also do things like going

in saying well what sorts of genes are
different expressed so one of the groups

of genes that's very differently
expressed are these small nuclear R.N.A.

genes there's a whole family of about
twenty of them fifteen or sixteen of which

are in the top one hundred differently
expressed in these snore days.

There's another family snore
a lot different to express so

I don't know what's going.

On With that they're not
in one place in the genome.

So there's some court regulation

there annotated as having involved
a role in epigenetic modification.

But that's about all on our.

Oxidative phosphorylation

proteins is certainly different
expressed between the different places.

So I don't know if that reflects
a differential rate of activation and

and maturation of immune cells but
it's certainly intriguing.

So since we have the gene
attacks what we can do

is do a genome wide association
study on gene expression variation.

OK So genome wide association
studies where you take genotypes So

in this case half a million of them so

five hundred thousand phenotypes
throughout the genome for

every one of the participants in the study
and at each genotype you ask the question

is the different associated with
the phenotype so you can do this for

diabetes or you can do it for
schizophrenia or you can do it for height.

OK So that's one trade at a time and
when you do that you're looking for snips

where you get a significant association
at the level of ten to the minus eight or

better indicating genome
wide significance.

So if you did so for
example this was done.

Has been done for serum leopards and
they found one hundred

variants genome wide significant
defect serum Lippa concentrations and

they explain about fifteen or
twenty percent of the genetic

variation serum Lipitor about
thirty percent of total variation.

OK you can do it for
diabetes there are about thirty known

genome wide significant associations
again that explain about fifteen twenty

percent of the genetic variance and
those studies are done on.

Upwards of five thousand cases and

controls in fact the serum liquids one is
done on one hundred fifty thousand people.

Here we're looking at two hundred people.

OK which you might say well that's that's.

That's not going to work but
in fact it works wonderfully.

And this is what we call a Manhattan plot
where we're walking along the genome.

So from the tip of the first
chromosome down to the.

Tip of the other tip of the twenty
twenty second chromosome.

Which chromosomes different color and
what you've got on the Y.

axis is called the negative logarithm
of the P. value of the association.

OK So people you tend to minus
eight has a negative log of eight.

So that's how cutoff.

So ninety nine point nine percent of
the Jena type associations are below

that cutoff.

But we have about fifteen hundred
associations that exceed that threshold

cutoff and there in four
hundred independent lowside OK.

So you see a big cluster here that's
actually on the on the six chromosomes

that's in fact the M.H.C.
cluster so there's a lot of there.

So what it's saying is that genotypes in
the M H C A regulating the expression of

genes in the M.H.C. cancer we've got about
four hundred independent associations and

it turns out that many of those replicate
associations that have been detected in if

you just look a little that's cell
lines that other people have done.

This plot here is plotting the location
of the snip along the chromosomes

against the location of the target
genes that it's associated with.

And what you see very clearly
is that almost all of those

associations along the diagonal.

And what that means is that
the snip is in the same location

of the genome as the target gene.

So in other words these a variance in
that gene to regulate its own expression.

So that promoted one of morphisms and
regulate engine expression and

the reason why this approach works so
well is because it turns out that promoted

variance will explain as much as twenty
percent in some cases up to fifty or

sixty percent of the variation in
transcript abundance for that gene.

OK.

So genetic influences on transcription
can be very very large they can

explain twenty or thirty percent of it but

by the time you get the genetic
influence on the disease itself.

It's diluted out by the environment by all
the complex interactions that are going

on and it's very rare for any genotype to
have an influence of more than about one

percent on the gene on the actual
trait you get in the end.

OK so the.

Why there's a lot of interest in
doing the genetics at this level.

So we can actually desexed genetic effects
regulate gene expression we can see

the large effects and study that much more
easily than genetics of disease per se.

And then there's about a dozen places
where you have what are called trans

associations so
these are snips in one chromosome

affecting expression on another
chromosome but they're quite rare.

And there are certainly some
interesting cases there that actually

capture No one disease as well.

OK but

out question was as I said was how robust
are those associations to the environment.

And I can show you three
hundred eighty of these plots

showing that they're very robust.

So what you're seeing here this
is arm the gene expression level

normalized on some scale so I actually
think about this scale as being arm.

So it's a lot based to scale so.

This is ranging over about a two
fold difference a minus a half

plus a half is about a two
fold difference in expression.

OK So this is the minor homicide.

The major harm a zygote so
that might be a on the zygotes and

these are the minor one is I got so
these might be the people who had G.D.

home as I gotz and then these
are the people who had to resign I guess.

OK.

And so you're saying that across
the entire population you see

you see that these minor homicide rates
have higher level of expression of this

particular gene in them than the Major OK.

But the other thing you see
here is that those lines.

The red blue and
green lines of Earth centrally parallel so

red is in fact a gram.

So the Arab village blue is blue
trough the burger village and

green is Agha DIA the city.

So as does Jean a typical FEX then
a constant in a different environments so

we know the environment affects
gene expression but at least these

cysts acting variants that affect
an expression completely robust to that.

And as I said I could show you three
hundred eighty of these these slides.

Here's a couple of cases where we actually
see that blue truck has difference of

expression relative to the other two
locations but nevertheless a gene to be.

Effect is constant.

This is a situation where it looks
like this crossing of line means OK so

in agood DHEA.

The hunt minor homicide lots
of higher expression but

in fact there's only two of
those sorts of rarely or so

the A little less than five percent
as only two individuals are rare and

it's just as likely that those two were
from these populations the other so

that's probably just
a statistical out of out.

OK it may be real but
it's more like statistical artifact.

OK So we do the analysis it turns out sort
of really contra to my expectations that.

There is no genotype by only
environment interaction at the level of

gene expression and that's a conundrum
because as I said at the beginning.

I'm interested in this question of why
it is that disease is increasing in

the modern world and there are some well
there's an interaction between genotype

and environment giving rise to increased
disease susceptibility and yet

when we look at the very
molecular level of gene or

try to fit into an expression
we don't see it.

So here's how and thinking about that.

So if we just think of one example.

OK so I've done.

Just this a different example but
again that suppose and I G.G.

Let's suppose that the people who are on
higher risk of some disease diabetes for

example are the ones who have very low
levels of gene expression because you

don't have enough transcript to to do the
job and you're a higher risk of disease.

Well the only people who
are really at higher risk

of disease as these ones down here.

So it's the minor homicide
goes in that population.

OK so
you can have an absence of interaction.

So what we're seeing here this is blue.

This is blue truck.

OK low expression intermediate
expression high expression

in all cases lower than you
see in a gram and Agathe here.

But if what affects disease
is gene expression.

Then the only ones at risk
are these individuals here.

So you could see an interaction
between genotype and disease.

Despite the absence of interaction
between genotype and gene expression.

OK so we're now sort of ag.

And looking forward to sort of do this
approach where we say well can we link.

The intermediate phenotypes to disease

through throughout you
know type associations.

So that that's one study giving
a sort of sense of what we do.

Here's another one motivated
by this idea that maternal

health during pregnancy in particular
can influence a child's development.

So the fifty phenotype hypothesis is
actually the observation that underweight

mothers tend to have children who are at
risk of early onset of the city and

diabetes and other chronic diseases.

And the argument is that if the mother
is experiencing poignant Trish and

then she will reprogram the child's.

Biochemistry to actually be able to
rapidly assimilate nutrients and

then when the child is born in a world
where there's there's a ready availability

of nutrients then then.

They got to put those in fact deposits and
you get you.

Simplistically speaking get early
onset diabetes and the sequel of that

I think more prevalent in modern society
is the impact of obesity and overweight.

On disease risk so we know that
there's heritability of obesity.

But we also know that we can tease that
out because you can look at the mothers

and the fathers influences and we actually
know that maternal obesity during

pregnancy is probably having
a direct impact on child obesity.

So we said can we detect that already at

birth by looking at gene
expression profiles.

Now obviously the only tissue you can
get from it from a newborn child is

cord blood.

OK so the cord blood is in origin.

So what we did is we set out to
compare cord blood to maternal blood

from matched mother child
pairs where we had about we

wanted to have about twenty normal weight
mothers about twenty obese mothers and

about twenty just ational diabetic
mothers of technical reasons we're doing

get that sample size we've got enough to
sort of start arm seeing some interesting

effect so and that study was done by you
sort of precisely Mason some of you may.

Remember she was here for
six months when I first moved and

then Graham Trant was the O.B.
G.Y.N. that we work with.

So in this case we saw a whole new axis
of variation that we didn't see it.

We haven't seen in general adult
population and so the first principle of

expression variation separates these
individuals from these individuals and

we don't know a whole lot
about what's different there.

I can tell you that one of
the differences is the ratio city for

a City eight to lymphocytes.

But that effect is at least as large as
the effect of mother versus child OK So

this is the mother's sin.

So the second prison term
is separating the mothers.

From the children.

And that's about two or three thousand
genes the differential expressed to

him other than child that was not on
replicated what other people are saying.

What's new is this observation of
these two classes of individuals and

we've got a large enough sample to
start seeing this differentiation.

If we go down further.

And the other thing I would should say is
that if the mother is on this side has

low P.C. one values in every case her
child had low P.C. want to values.

So there's perfect transmission of
whatever's going on in the mother

to her child.

We presume probably through cytokine
signaling influencing both maternal and

the fetal gene expression.

Now not all of those acts is a remotely
associated with with obesity or other

obvious measures of health but if we go
down further we see something interesting.

So the third Prince component axis is
in fact equivalent to the first axis in

the general adult population but the
fourth axis here is what I want to look at

right now because the blue spots almost
all of these are the children born to

obese mothers and the pink spots are the
profiles of the obese mothers themselves.

So you get a very clear separation which
is significant at nominal sort of point or

one level and then the children
born to just a diabetic mothers

are also on that side of the equation.

So we're going to try to sort of
replicate this here in Atlanta but

it's our first indication.

Then possibly there's an impact on
maternal health the regular gene

expression profiles for several hundred
genes which are in fact in rich for

genes involved in central metabolism.

We can do some network analyses of these
so this is Bruce our now has a program or

set of suite of software he's written
called Top gene which starts on a gene

ontology analysis but immediately outputs
it into side escape network analyses So

what you're seeing here is all of the
differential expressed genes between the.

The type A and type B.
maternal fetal profile.

With their network to regulate it.

So these pink squares here are in
fact transcription factors so

all of these genes are targets of.

These transcription factors.

And then some of these the darker
purple ones are micro R.N.A.'s as well.

So these are potential targets
of those micro are nice.

So you see a regulated response so
these are our candidates for

the mediating that difference.

The brown ones here are all genes
involved in chromosome modifications

I think there's a little color quadriceps
down here there's some I think there's

comment in modifying factors here to them
what's interesting to us really is this

yellow once these are all genes that
are annotated as being responsive to meet

the trackside which is a drug that mothers
take to prevent spine of spinal bifida.

During birth and
it's also a drug that taken.

In various other contexts.

So.

You know what are the potential long
term consequences that we don't know but

I think it's pretty important to
sort of consider that maybe people

are differentiating their response
response to these drugs and

that's significant at ten of them on
a seven or eight level that association.

So we can do a sort of network analysis
the other thing we can do is to hone in

on specific pathways.

OK So this is just a quick
look at insulin signalling and

what we can do here is
just cluster the people.

So each column now is one of
the individuals in each row is a gene.

Involved inch and signaling and

you can see that some of the genes
have high levels expression.

Another one's lower levels expression and
then they switch to

different components of the pathway so the
grain in the blue axes here are different.

With respect to these disorganization So
I'm actually teaming up with Melissa and

hang low to now start seeing if we
can use microfluidic our SES and

other things to really hone in on
particular pathways and link the variation

we observe at the population level to real
functional ASOS where you can go in a sort

of knockout or what Melissa does
knocking out individual genes or

doing specific casts ace and really tie
the variability in the population level

to what individual genes are doing and
I'm certainly really keen

to work with different people in
engineering applications here and

others to really see if we can add
a functionally to what's going on.

OK so we can measure gene expression
profiles metabolite profiles

we're also doing.

We can measure clinical covariance but
I think what we want to be able to do is

also go in and measure actual functional
aspects of of these cell lines.

I'm stretching this a little bit because
this is actually not that not that pathway

but one of the components a variation or
so varied by location in Britain.

So you can see here that are the blue
dots are all over the city but

there's an absence of red dots in the
center of the city and this is actually

a height the sort of high income high
socioeconomic group region of the city and

we're interested to know
whether that sort of thing.

Also replicates here in Atlanta.

So one of the things we can
do is actually take people's.

Zip codes and sort of sort of.

Two geographic analysis of that.

So we think they're actually not just
that you're seeing sort of rural

villages versus cities but
we think there might be fine structure

to variability across even
large cities these days and

they certainly have some interest in
treating literature emerging about.

So she cannot make an impact on
chronic disease and gene expression.

So conclusions for that sort of work.

Gene expression in human blood we
know it's highly heritable but

it's also extremely.

Strongly affected by the environment and
what I'm interested in is it does to

sect in the impacts of geography
gender culture and so forth on that I

was surprised not to see energy but
type by environment interactions in this.

But as I mentioned.

What's interesting is that additive
affects the expression level so

that's what observe
might necessarily might.

Nevertheless translate into interaction
effects of the disease level.

And the idea is in that systems genetics
is merging genetics with endo phenotype

such gene expression hormone levels
metabolite abundance and so forth and

we're opening up this black box
between genotype and disease.

So instead of just doing genotypes disease
we're doing it in size with these and

of in a types.

So that's all well and good so
it's sort of very arm.

What is a terrorist work or
sort of basic biology but

what we want to do in the end is actually
bring back these sorts of studies

to disease and as I mentioned the
beginning of an evolutionary background.

So I've been thinking about these
things in flies for a long time.

ANNOUNCER trying to transport that way
of thinking over to human genetics.

And here's the question so why is it
that in disease incidence is increasing.

OK well.

We're all familiar with Bell Curves in
this case what we're what we're looking at

is just.

Said to be ready for some disease
might be diabetes might schizophrenia.

OK so.

Most people are going to have
an intermediate number of

susceptibility ills.

OK.
Some people by chance you've got to have

more illegals increasing risk or not some
people don't have morals decreasing risk.

And if we look through our human history
systems evolve to make sure that

susceptibility is decreased.

It is that you minimize
settler to disease.

So this is our point you don't have one or
two percent

above the threshold of susceptibility
illegals giving rise to disease and

we now know that for most diseases
the number of illegals that are involved

are in the hundreds if not thousands.

OK so it's a very small genetic effects
but nevertheless systems of evolved.

Vente disease and.

What most people think
about what's happening now

with the environment changing our
susceptibility is that you're actually

shifting the susceptibility a little come.

OK Well our genomes are involving in
the space of a few generations that

there's not enough time for that.

So somehow the interaction between
types and whatever the environment is

it is increasing the number of those
that are actually susceptibility ills.

OK So genetic effects which in the past
were not sufficient to give rise to

disease in the modern environment are
sufficient and there are more of them and

so they're actually shifting
your whole distribution over.

The other possibility which I get from
my background in flies is that in fact

we're seeing something
called De canalization So

instead of getting a global
shift of the profile.

We're getting an increase in variance
of susceptibility really all effects

and the reason why you have ten or
fifteen percent of people at risk.

Now is because of this increase in
variance not so much as a global shift.

So that's that sources that we're seeing I
won't carry on more about it for today's

talk but it sort of gives a background to
what is that we're now actually starting

to focus on disease cohorts which brings
me to the what we're doing here and

the role of the center in the political
institute and what's going on.

So I mean how many of you where we
actually have a predictive health

institute here at all.

Bob thanks Bob just a few All right so

it's a joint Georgia Tech Emory
initiative focusing on health.

OK which is combining the strengths of
both universities so if we think about

those then we have enormous strength in
bioengineering here but both universities

actually were particularly IB so I'm
excited to sort of be part of this group.

We have enormous strength about for

Magic's a particularly strong in
comparative genomics sequence analysis and

increasingly now and
in quantitative genetics.

We have enormous strength in things
like health care economics in database

management and so forth and so
one of the people I'm working with now is

Nicollette a Serb an industrial so
systems engineering and just terrific.

Integrate across what so
many disciplines and units here.

And then of course Emory has has a huge
number of clinical cohorts they have one

of the strongest human genetics
departments in the country and

they have access or

right there is the Children's Healthcare
of Atlanta with its focus on pediatrics.

So we can put these things together and

I think it's really an underutilized
at this point as to what much we're

doing that now one of the center pieces
of the predictive health institute.

So there's some about three or

four five million dollars to put into
this by Emory for five years ago and

what can Brigham has done is to set up
the Center for the health discovery and

wellbeing which is located up here at
Midtown hospital up on Peachtree Street.

And I'll tell you about
that cohort in a minute.

But as I said I think there's
an unrealized potential for

Georgia Tech involvement and one of the
other things I did say in here that I'd

like to see happen is that this is to have
a spring in environmental bio sensing

because the places that are doing genome
wide association studies really seriously

you know the Broads and the and
the Welcome Trust and others and

pumping hundreds of millions of dollars
into that I mean they're really looking

at hundred thousand people at once
we've got a cohort of six hundred.

OK.

So we're going to be a little bit
flexible about how we approach this.

But we can do that and one of the ways we
can do it is by being very clever about

engineering applications and and so

so doing things such as I have been
doing with Melissa and hang but

also I think the environmental component
people are just playing lip service

to we know the environment you put
we don't go out and measure it.

So I'm really hoping that maybe
this mention is in the audience

here that we can do things like
putting sensors in homes or

attaching them to people's clothes and
measuring how much exposure to to P.C.

be and to other toxins in the environment
they've got lead and so forth and

we can integrate that
into our into our study.

So the C.S.T. W.B. cohort is about
six hundred Emory employees or

or friends of the university in the.

Age range of about twenty
five to seventy five.

OK so they're healthy at that residence.

OK but healthy is a relative term.

So we've got about thirty percent.

Obese in the study cohort the high
risk of metabolic syndrome so

a lot of these people in fact most
of them or older than forty and

many will be coming into very high risk of
cardiovascular illness and diabetes and so

forth in the next five to ten years so
it's a really good time to get this cohort

and what is happening is that they're
coming in and getting a very very

intensive clinical profiling done which
takes about five or six hours of tests.

So they're getting their blood cells drawn
and getting complete blood cell counts and

immune profiles they're
getting cardiovascular tests.

They're getting psychological evaluation
bone mineral density going through all

sorts of scanners and so
that happens every six months and

then they're partnered
with a health partner.

OK So this is someone who talks to them
on a regular basis about the dire about

their exercise regime and about reducing
stress in their lives and the objective is

is to sort of change the way we do
health care put the focus on wellness.

If you could reduce the rate
of chronic disease.

OK so in America.

One of the major diseases to
cardiovascular is it's depression.

It's diabetes.

Metabolic disease.

If we can reduce those through the simple
lifestyle interventions such as this which

is about two or three thousand dollars a
year the health care savings can be really

quite substantial.

So it's a pilot to see if that works and

what we've got the opportunity to do
is to go in and do to nomics on that.

OK so with Dean Jones over
Damrey he's doing for you.

Transform aspects crop take spectroscopy.

So he's got three thousand metabolites
that he feels he can robustly measure.

So for the first time we actually
will have a Cold War We've

got the whole genome
phenotypes the metabolites and

the complete transcript and we've done
now we can put those things together and

try to integrate those of course with all
of this wonderful clinical data as well.

So we've just got that data random which.

Starting to go through it now and
then as I said we're actually working on

functional assets as well the most
interesting finding that's come out so

far is in fact not from this cohort but
it's with an affiliated one which is our.

Coronary artery disease cohort so he's
being people into his clinic who get chest

pain or more advanced coronary disease and
so we've profiled two hundred of those and

done the gene types on
those as well in parallel.

So there's sort of a ten year advance
cohort over the C H D W P one and

this is sort of new data.

I'll come back to that
since I've introduced it

which is this is one thing that
I'm really excited about so.

We've got a profile of about five hundred
genes to second principle component of

variation in this cohort which is very
strongly associated with death by my

count and function function so that
the people who actually in that cohort who

subsequently died of a heart attack
if they have this profile here.

OK then they're pretty much unlikely
to have died of a heart attack.

Whereas they had that profile here and

they got much much elevated
risk of death by Emma.

So that's the sort of thing that we're
really interested in going in now and

looking at variability in that and
we want to know if this changes over time.

So one of the other
nice things about that.

C.S.T. W.B. cohort is we're getting
profiles every six months of

the people and we'd like to see if those
interventions can reduce that risk

take people out of this profile
down to this sort of profile and

then of course we'd like to go in and
sort of see what the genes are there.

But the other thing that sort of
strikes me in this cohort is just

the repeatability of our measures.

So this is triglycerides measured three
times in each of one hundred fifty people.

OK So there are some people
who've got extremely high levels

of triglycerides at enrollment and
six months later.

This is insulin.

Same deal extremely high levels.

So one thing we can do now is actually go
in and sequence those extreme individuals

and see if we can find variance by
sort of whole Exxon sequencing and

others that might be responsible for
these.

Affects but even H.D.L. major predictor
of heart disease the repeatability

this is absolutely astonishing to me this
is the significance of that correlation

three repeat measures of one hundred
people is about ten to the minus ninety.

OK.

So people there.

So how able are we to get people
from in this case low H.D.L. is

not good up into this range by lifestyle
inventions that's one of our questions.

OK And then we've got this rich cohort and
we can do things like asking

how to genes affect simple
tenuously triglycerides and H.D.L.

So here you see repeat we can repeat that
measure that that we know that people with

higher triglycerides and low H.D.L.
are at higher risk and so there they are.

So that's what I want to
be doing in that cohort.

So other things we're doing so

there's a cystic fibrosis project with
Nama Carty here and collaborator.

So we're asking the question Do airway and
peripheral and opera for

blood neutrophils display a particular
profile of oxidative likely to express.

The might differentiate the C.F.

patients who go on to get
diabetes from those who don't.

Or does differentiate all of these people
with that chronic stress and the way for

normal people and

what we learn about the stress in that
system where working with an Merton's

on a childhood leukemia survive a cohort
let's mention some people in lab earlier.

So here the question is to steroids and
radiation therapy

every genetically modified gene expression
and is that related to the onset of

chronic diseases like bone mineral density
loss in that cohort which is a very

very serious problem for
early childhood cancer survivors.

Again the connection with the Children's
Healthcare of Atlanta is fantastic.

Just recently I started working with a son
who is one of the world's leaders in

pediatric Crohn's disease so he's got
a cohort there he's leading an effort to

assemble could actually Crohn's data so
he's getting so close.

He's an inflammatory bowel disease so

he's getting biopsies she's
getting the micro biome I.

Microbial profiles he's getting
peripheral blood all of these things

we can put those together and

ask whether a combination in this case
our first question is now we want to know

whether Gina types can predict severity of
disease and in fact the need for surgery.

OK so it looks like that might be
the case with some of the hits but

outthought is that it combination of types
of gene expression is actually going

to improve that to the point where
it becomes clinically relevant So

that's the sort of thing we're doing
there and then I think there's

hope there are some people in the audience
interested in stem cell work.

So we're actually have a grant team with
hange and young sort to to our genome wide

association studies on variable I.P.S.
cells profiles with the interest

there of NIH how reprogramming is
affecting individuality of profiles

as we go ahead tick in the context of
cardiovascular disease in this case.

And then up some you may have seen
the other sort of fun project I have

on the side is the facial imaging
projects who are actually interested in

the genetics of expression
of emotional expression.

So we're looking at that in adolescence
but we're also looking at in children and.

We're going to ask the question whether we
can find variants in particular genes that

are associated with development of
facial morphology the going here.

We're really interested in as well
as looking at the genetics of

the emergence of
the expression of emotions.

You know Darwin wrote the expression
of imagine of an emotion in Man and

Animals which is the great
treaties written.

You know hundred years ago and
there's almost no follow up not on that.

So in the long term that's actually
something I'm very excited about.

So just to start to conclude then so that
the Center for Integrative genomics is

a research class to here that I'm putting
together dedicated to the notion of

systems genetics approaches can eliminate
the molecular basis of disease separately.

But we're doing out in the context of my
interest in the evolution of disease and

what I really want to do is to integrate
what we're doing with the bioengineering

capacity and

bioinformatics capacity that we have here
to really make sure that Georgia Tech

is playing a major role in the predictive
Health Institute as we move forward.

And I think my last slide is a and
acknowledgement slide

this picture is actually from Youssef
Carson who won the UNESCO prize for

the best photo this year and so

you see a little girl being taught to read
in one of those villages that I mentioned.

So you said something to doll I was
at the International Foundation for

conservation and development of
wildlife which is an institute devoted

to ecological research that
supported us in Morocco.

Ross Wendy and Kelsey at the SAS Institute
within long time collaborate with them.

David and Kevin at Duke and then Peter
Vischer is a great statistician you cue.

Basement study was done actually profound
an interesting well is this group

by Lucy Mason I mentioned grammar.

I was working with a whole lot
of people here in Atlanta.

So I cannot shed and others can he
is a graduate student starting to

work on this here and then I also really
want to Arafat is my lab manager here.

Thank you thank you.

This is Bob Goldberg.

Yeah.

So this is this is pat downs I got so
that was great.

My question is I don't like
gene expression she planned for

a while I could tell you if you're a
graduate student back to take an exam it's

going to change the question and
very very much.

But if you marry.

Well yeah

that's so that's what I'd like to know
I'd really like to know that it's to

small a couple to get that we only
have about a half three or four.

How can you be in prison straight or so.

So you can start to be bitter.

I think of business Britain won the award
for the most of the city in the world.

Last year that you wouldn't notice in
the city must that there's actually no

correlation with obesity and that does
profiles there which surprised me

but yeah we need a bigger study.

So in fact I'm really hoping to what was
some of the pediatricians over at Emory

to do a replication of this in Grady and
call for long but it's just takes quite

a bit of effort to get the surgeons in
line to be prepared to take those samples.

Good question.

Now there's an opportunity to kind
of experiment where there's a major

population is dispersed changing
their environment without changing

or people leave after
the earthquake is right.

It's kind of an experimental.

Separate from there.

Yeah I think it would be wonderful.

I mean my dream would be to
sort of go around the world and

do a map of expression profiles
how they vary but by geography.

We're getting a hint.

Now in Atlanta that there might be
a difference between ethnicities.

But it's hard to tease that apart from
socio economic things I want to do that.

The problem is it's incredibly expensive
to actually assemble those cohorts

just just in personnel to go and find and
get those and you're not going to

get funding to do this until you've
got the studies in the freezer.

So that's hot.

So instead I think the closest to that
will be looking at migrant populations.

So there's a very large so
I should actually as

a service where he has the good population
here in Atlanta this is a very large.

North South Asian population.

OK A lot of interest in that.

So they're actually we have
health as with those people.

So we're starting to profile their blood
because population we can look at we have

clinical data around and they.

Compare that to what's going on in India.

It's on our soil working with Anka and
Ryan.

You may know it in public health.

So they've got a four thousand
person study going on in Calcutta.

Because in India something so
fascinating in India arm.

You've got now some of the highest
rates of diabetes in the world but

it's not associated with with obesity
in fact my so one of the students

go presentation in my group last week and
sort of showed some genetic data and

she said well if I controlled for
obesity the signal went up and

said well that's really interesting
how do you define obesity she said.

B.M.I. above twenty five.

I was like damn I just lost twenty
pounds and I'm at twenty five and

now you tell me I'm still obese.

But that there's real and then I'm looking
in Fiji as well about a student in

Fiji doing things where you've got
actually the Melanesian population and

the and the SO Fiji is about half
Melanesian half south Asian or

Indian because overall brought
in what sugarcane plantations.

What in the latter part of the eighteenth
century ninety eight in central

obesity rates in both are through
the roof thirty percent.

But with dip totally different obesity
profiles you know South Pacific Islanders

have had obesity for as far as long
as you know if you look up and

quote the she was there but
they didn't have diabetes.

So there's probably associated with
a shift in the sort of oils of people cook

with or things like processed foods so
so you know the villages will now

go out to sort of buy grooves and improve
their sort of living standards they will

actually sell their healthy foods and
then what they've got left the food they

have to buy process canned stuff and so
their diet is just going is terrible.

So you've gone in one generation this
demographic transition to very high

levels of cardiovascular disease
hypertension and diabetes.

So we'd love to study things in
those populations where we've got

collaborating people.

She said you talked about.

Catch up to them and there's where is that
verse some of those developing countries.

They have evolutionary conditions.

Now everything is better nutrition so
yeah.

So I look at it as that.

So what we call advance
physiologically is is is not about us.

It's the other way around.

You know.

In any

way I

mean and

you're sounding like Jim Nill I
mean that's the real sort of

stuff the guy doing nearly you how do
you heard of the fifty genotype type

idea was that the reason we're
getting diabetes is because there was.

Humans doing evolution recent
evolution passed went through

massive periods of famine so there
would have been a ten ten selection for

the ability to store a couple hydrates
rapidly and now that we've got ready

availability of that those genes that
will good then and now bad for us.

So it's sort of a classical fundamentalist

Darwinian near the end paradigm and
it's almost certainly completely wrong.

OK So all of the data now show
that we know the thirty or so

diabetes set really last I none
of them have a signature of

of recent selection on them
where I'm going with this.

Yeah not so I know he's not the guy
who wrote the paper so so did.

Neil put this idea out there and it was
forty years ago or so instinct idea.

Turns out to be wrong but

in the paper he said if it should turn out
that in fact fifty genotypes are true then

we need to do what we can to preserve
them in the human genome and you sort of

thinking well how you gonna do that when
you tell this person that you want to mate

with that it's kind of scary that
there was a lot of eugenics back then.

So you're saying can we create
a perfect human environment.

I think what we've got to do is realize
that we're all genetically very different

and what's a good environment for you
is not necessarily good environment for

me and that's something where actually so
I've also got a fly group and

what we're doing is growing flies on four
different diets a high fat diet a low

sugar diet a high sugar diet and across
the one hundred fifty lines we've looked

at there's actually no impact on weight
gain or other phenotypes but the different

diets do complete different lines to
completely differently on different diets.

So what's good for this line on what this
one does best on high fat This one is

going to do best on high sugar and I think
it's almost certainly true of us as well.

If you read Michael Pollan's work
I don't know many of you know him.

The guy's written on the biology
of desire or in defense of food.

He sort of makes that point that we've all
got if we can't reduce things too much to

sort of saying that this is the optimal
type the take on message is variability.

We're all variable and what's good for you
and good for me is not actually good for

someone else we've all got to find our
own right place with respect to that so

that be my answer in a roundabout way.

Volatile.

OK for me.