Hello everyone.
Welcome to today's webinar, which is the final webinar of our summer omics
learning seminar series. The series is presented
by the omics and bioinformatics core of the M-LEEaD center at the University
of Michigan. M-LEEaD is the michigan life stage
environmental exposures and disease center. In case you
missed our previous events, the recordings are
archived on our website and I'll post the link to that website into the chat
shortly. Before we get started I'll just orient
you to the zoom system, although I know many of you are
familiar at this point. Attendees are all muted upon entry and
for the duration of the event. Please submit questions for our speaker
Dr. Wright into the Q&A feature which you will see
at the bottom of your screen. You're welcome to submit questions at
any time during the event. Also this event is being recorded and we
will add it to our center website in the next few days.
In the interest of time I'd like to hand it over to Dr. Maureen Sartor
co-lead of our core. Hello welcome everyone to our third
and final zoom seminar of our summer omics
learning seminar series.

As Katie said the series is being hosted
by the M-LEEaD center of University of Michigan.
I'm Maureen Sartor lead of the omics and bioinformatics core
of the M-LEEaD center and I am extremely pleased to be able to introduce
Dr. Robert Wright from Mount Sinai in New York to you today.
Dr. Wright should feel at home with us here today since he grew up
in Melbournedale, Michigan right next to my hometown of Dearborn, Michigan
and he attended the University of Michigan Dearborn as an undergraduate
and then the University of Michigan Ann Arbor for
medical school. He's currently the Ethel H Wise chair of the department of
environmental medicine and public health at the Icahn School of
Medicine at Mount Sinai. He trained in pediatrics
at Northwestern University and medical toxicology genetics and
environmental epidemiology at Harvard.

He's the founder of Mount Sinai
Institute for Exposomic Research in the Lautenberg
Laboratory of Environmental Health Science,
which specializes in biomarkers of environmental exposure and effects.
His research is dedicated to studying the effects of early life
environment and exposomics, which includes
nutrition, social factors, toxic chemicals, and the physical environment,
including air pollution, green space, etc. And he studies this on the effects of
these on child health and development. He's also PI of and founder of an
ongoing birth cohort study funded by the NIH called PROGRESS which
stands for programming research in obesity growth environment and social
stress, which is based in Mexico City and is designed to address
complex mixtures of chemical exposures as predictors of neurodevelopment and
the role of psychosocial factors as modifiers of chemical toxicity.
Bob has published over 300 research papers,
most of which deal with environmental health issues,
and has served on numerous national committee and advisory boards
in the field of pediatric environmental health.
So today he'll be speaking on "Exposomics and precision medicine –
everything that rises must converge." So I'm interested to hear
what he means by that so um and Bob, I have to give you an extra special thanks
for presenting with us today especially since
you're taking this time out of your vacation week
so we're um we're very grateful for that and I'll go ahead and turn it over to
you now.

Thanks thank you Maureen can everyone
hear me? Yes. Great so uh for those of you not
familiar with Mount Sinai this is a photo of us. That tall building is
part of Mount Sinai and is at 102nd street
and our entire campus goes from 98 to 102nd street so four city blocks and
it includes fifth avenue and madison avenue fifth avenue
is this avenue right across the street from central park
and this is madison avenue on the left side so that's our entire campus.

as you can see very much a vertically oriented campus
and our laboratory is in this building here if you can see my arrow and my
office is excuse me in this building. So I'm going to talk today about
exposomics and also the field of precision medicine
which I think have been rising in parallel and I think
uh as the subtitle um suggests they must converge so this is actually
uh the title of a Flannery O'Connor short story
uh for those of you who studied uh southern literature from the 1950s
and 60s Flannery O'Connor was a famous author who in this
particular story was talking about uh white poverty and black poverty uh
converging um perhaps ironically given the times
but the title that uh the reason why I chose
it is because I believe exposomics is rising as well as precision medicine
and I believe it's time for them to converge.

So and as an outline of the talk I'm
going to first talk about how we can apply exposomics to understand
health and disease. I'll talk about what precision medicine is,
uh why I think it is the future and I will note that precision medicine is
largely ignoring environmental health and
exposomics but I'm going to also make the counter argument that environmental
health and exposure exposomics has largely been ignoring
precision medicine and it's really time for the two fields
to work together. Otherwise it's time to converge.
So what is precision medicine? So the national research council definition is
this. It's the tailoring of medical treatment to
patient characteristics so that we can create susceptibility
classifications through disease, it's prognosis
or a patient's response to treatment.

By doing so we can individualize the
interventions that we provide, maximize benefit, reduce treatment
failures and reduce side effects. Fundamentally what it really means is
understanding and applying each person's individual background and how that
background influences disease progression and
treatment response. And this is an important distinction
from what we typically do in environmental health. Precision medicine
operates in a setting where the probability of illness is one.
Prevention no longer matters to a physician because
for a physician taking care of patients, particularly subspecialists, the patient
is sick. How they got there is actually less
relevant to what he or she will do for that patient
than how the environment currently is for that patient.

So in other words their
past environment that may have gotten them there
is not as important as their current environment. And we tend to focus on the
past environment. So that's precision medicine, so what is
exposomics? Well it's not just chemicals, it's a
study of all the health relevant environmental
factors across the lifespan. And it's obviously gigantic in scope and
in fact it's bigger than the genome. The genome is 4
billion base pairs. The exposome is nearly infinite, so
part of the problem with exposomics is when you're getting a working definition
and getting your head around what can and can't be measured. I prefer this
definition and it's all the health relevant environment across the lifespan, not all of the environment just those important factors that have
greatly influenced our health.

And there's many of them that we can measure
and in fact every year we get closer and closer to
measuring the exposome. Because Iggy Pop grew up in Ann Arbor, I
thought I would quote him. It's very hard to find an Iggy Pop quote
that's actually uh something you can say in public but this is one: It's very
important to know what not to do. In other words you can't
measure everything so we have to understand what we can measure
and you have to focus on the things that are important.
So right now the field of precision medicine is practiced as a
pharmacogenomic field and the idea is if we understand genetic
variation to disease and treatment then we can target therapies the most
effective drugs, improve efficacy, decrease toxicity,
and target behavioral modification.

Somebody's at higher risk because of
their particular genetics.
This way we can improve counseling and decision making, improve
patient outcomes and satisfaction. However it really would only work if
genes were the only source of variation in treatment and prognosis.
And in fact if you look at websites for precision medicine at different
universities, and none of these are University of Michigan but these are
actual websites, almost all of them talk about genes causing disease and
genes being used to target drugs to treat a disease
and they all have schematics that show DNA in one way or another. And almost none of them mention
environment but they all have these DNA and these pills and stratification of
risk figures in them. Sometimes they're
oriented horizontally, this one's orient oriented vertically, but it's basically
the same idea. We know that genetics is only one piece
of a much much bigger puzzle yet we keep focusing on the genetics and
ignoring the environmental components that are clearly playing a role not just
in disease causation but in response to treatment.
And we know that genes don't cause all of complex diseases and we've known this
for a very long time, in fact this is the cover of Science in April of 2002
and it talks about complex diseases as a puzzle and it talks about all the
different pieces that make up that puzzle, it includes genes but it
also includes behavior, and nutrition, and infections, chemicals,
the physical environment, culture, society and stress and in fact
the introductory editorial is actually entitled it's not just the genes,
so despite this knowledge it goes back almost 20 years at least if not longer
we tend to focus almost solely on genetics.

So how did we get here? So complex
disease research focuses on complex diseases and these
are the diseases that drive health care costs in developed countries across the
world. They include things like ADHD, obesity,
asthma, COPD, renal failure, Parkinson's disease,
Alzheimer's disease, cancer and many other very prevalent diseases
that are very difficult to prevent in many cases very difficult to treat. We know that their etiology is a mixture
of genetic and environmental risk factors. We also know that in many cases these diseases are getting more and more
prevalent so they have higher prevalence, higher
annual incidence – diseases like asthma, diseases like
obesity. You know autism is kind of controversial whether or not
it's due to increased surveillance or to actual actual increases but obesity is
pretty easy or pretty objectively measured and that's clearly getting more
and more common, as is asthma and many many other diseases and certain forms of
cancer becoming more and more common.

So the prevalence in those cases is
indisputably rising. If it's indisputably rising, that's not
genetics. Genetic main effects cannot explain
rising rates of diseases that occur over a time scale of 10 to 30 years. Genetics needs multiple generations to change
the prevalence of a disease, plus it would be counterintuitive that diseases
that reduce the probability that you're going to reproduce would be selected,
that's the whole basis of natural selection. So
if anything these diseases shouldn't be increasing, they should be decreasing
if they are purely due to genetics. Unfortunately the environmental
components of these diseases are largely unidentified. I would argue in part because we focus a lot on genetics and we haven't yet
focused on the environment. I think exposomics is going to start hopefully
helping us understand the environmental causes.

I think that's because part of our language also creates
barriers. We tend to think of nature versus nurture and even in the way
we talk about diseases in their causation, we do heritability estimates. Heritability estimates you know claim that they that a particular disease is x
percent genes and x percent environment. It ignores the probability
that or excuse me ignores the fact that
genetics in most cases confers risk and not causation, just as environment
confirms confers risk and not causation. You carry the APOE4 variant, you're not
destined to get Alzheimer's disease. It's a probability. The probability is
higher yes but it's by no means a hundred percent. Even if you live to be a
hundred. And the levels of risk vary depending on
the prevalence of the environmental factor
and the prevalence of the genetic factor and in fact that probability is
actually the probability of those two factors coming together
and that's why they are a risk and not a cause. So why don't we take a step back and think
about why diseases are called genetic or

So if all diseases are due to
interactions between genes and environment, I actually believe that
that's true, then if the genetic polymorphism is low
in prevalence in a population and the environmental factor is high in
prevalence, to us that disease is going to appear to
be genetic. And conversely if the environmental
factor is very low in prevalence in a population
and the genetic polymorphism is very high in prevalence,
that disease is going to appear to be environmental to an outside observer. This is a concept that's very
difficult to see in people but is actually easy to illustrate in chickens. And there really is a trait in chickens called yellow shanks. Yellow shanks is a discoloration of chicken legs with a particular genetic
variant if those chickens are fed yellow corn as
opposed to white corn. This is an example I once saw Ken
Rothman give many years ago. The genetic component follows an
autosomal dominant inheritance pattern. So imagine two farmers: farmer Jones and
farmer Smith. They live down the road from each other.

Farmer Jones has an inbred chicken flock. All of those chickens carry the variant
for yellow shanks. for generations his family has always
fed their chickens white corn because yellow shanks reduces the price
at market. One day the store runs out of white corn,
he has to give him yellow corn. All of his chickens
get yellow shanks. If you were to ask farmer Jones what the cause of yellow
shanks is, he would tell you it's yellow corn.
The yellow corn is an environment or yellow shanks
is an environmental trait brought on by yellow
corn. Farmer Smith on the other hand also has an inbred chicken flock. None of them carry the variant for yellow shanks. Yellow corn is a little
bit cheaper so those families always bought yellow corn. And then one day a chicken gets a new mutation
and he notices that half of the offspring of that chicken
also get yellow shanks and then he draws a pedigree because he's an
amateur geneticist and in multiple generations he sees
that the pattern is at half of the offspring get yellow shanks.

So from his perspective farmer Smith would tell you that yellow shanks
is a genetic disease that follows an autosomal dominant
inheritance pattern. So which is it? Is yellow shanks, because
biologically it's the same disease and farmer Jones' chicken as farmers
Smith's chicken, Is it environmental or is it genetic? By the same token if
you think similar things are happening in cancer,
is cancer environmental or genetic if it's due to gene
environment interactions? And even diseases that we traditionally
think of as being purely genetic actually are due to gene environment
interactions. Phenylketonuria is an excellent example of this. You can
actually avoid the phenotype if you avoid phenylalanine early in life. That's even why we do genetic testing for it
and there's lots of genetic diseases for which it's actually easy
to talk about what the environmental factor is and they are also due to
gene environment interactions.

And in this context what is the
heritability of chicken shanks? Is it zero percent, which is what farmer
Jones is experiencing, or is it a hundred percent, which is what
farmer Smith is experiencing. Or maybe heritability is almost always
contextual and so it's actually not a real thing. It's something that we're using to avoid measuring what is
really happening environmentally which are gene
environment interactions, because all genes operate in variable
environmental backgrounds. And there's a lot of backgrounds. Nutrition is a background. Whether somebody is obese or thin,
whether they have vitamin deficiencies or mineral deficiencies,
that can affect the way metabolize drugs and also
affect their uh risk for disease. When you're pregnant, you have
huge physiologic changes which can affect drug metabolism. Geography
can also play a big role. Temperature can vary by geography. Allergen prevalence
can vary by geography. Altitude can actually increase your risk of
stroke because if you live in a very very high altitude
you will tend to have higher hemoglobin.

If you have higher hemoglobin, you have
more red blood cells, and there's a slight increase in viscosity so you can
actually have higher risk of stroke which has been
shown. Culture plays a big role in the in your lifestyle and in your
diet which and also your social systems all of which have very profound effects
on your health. People are not exposed to single
chemicals they're exposed to mixtures of chemicals and those chemicals vary by
geography as well as by culture uh and even things like
pregnancy can actually vary the number of different types of chemicals that
you're exposed to.

And perhaps the most important
background is health. If you are sick it actually changes the
way you metabolize drugs. A great example of this is back
when I was an intern, we used to give theophylline for asthma. It's not given anymore um but you know when kids had asthma
attacks often times it was due to influenza,
influenza could be a trigger. Turns out influenza infections actually
alter the way you metabolize theophylline and kids who
had an asthma attack due to influenza would
actually get theophylline toxicity because they wouldn't be able to metabolize the
drug as well if they actually had influenza
as well. It also increases exposure to multiple
drugs or what is often called polypharmacy uh if you're sick.

So health may actually be the most important background
when we talk about precision medicine and it's something to think about as we
try to integrate precision medicine and
exposomes. So here's some examples, so the classic
precision medicine example is this, a 53 year old develops a deep venous
thrombosis, a blood clot in other words, in her
upper leg on a trans-pacific flight. She's started out warfarin, which is a
blood thinner, and she develops a GI bleed because she was given a standard
dose. Turns out she has a genetic variant
which makes her a slow metabolizer of warfarin. She should have gotten a lower dose and if they had done a genomic screen they
would have found that and that is a classic
example of how precision medicine genomics can intersect and it is a real
thing and this is actually something that
could benefit patients.

But what about this example, a 17 year
old child with autism has increases of head bagging and anger
outbursts and if you were to do a genomic screen,
and autism is sometimes thought to be a purely genetic disease,
you would find nothing that would help you treat this patient.
This is actual case that I had many years ago. Turned out he had lead poisoning. His
blood lead was 73 which is very very high. A couple of things were unusual about
him. First he's 17 years old. He's actually the first patient I'd ever
seen that old with lead poisoning. Lead poisoning typically
occurs in toddlers two to three years of age. The difference was he had autism. Kids with autism have pica, they eat non-food
substances. It turned out he was eating newspapers
and this was the 1990s nobody has newspapers anymore but he used to eat
newspapers and the ink in the newspapers had a little bit of lead but he was
eating enough of it that he was getting lead poisoning.

And his three-year-old brother was not lead poisoned, because his three-year-old
brother did not have autism and was not eating the newspapers, so
his house was not the risk which is the traditional risk factor, it was actually
his autism that was a risk. And by understanding
that intersection between autism and lead poisoning we measured
his blood lead because of the pica issue and we were
able to treat him and his head banging anger outputs reduced. This is an example
where environment is actually changing the disease not
genomics and an environmental screen would have
potentially detected it, in fact that is how it was
detected. And here's another example, a 14 year old has gastroesophageal
reflux, a history of depression, and is treated with morphine for a
fracture and develops respiratory failure.

So genome scan revealed no functional variants in phase 1 or phase 2
P450 genes or phase 2 liver enzyme genes that could affect morphine metabolism
which is something that could have explained this. Turns out it GE reflux is being treated with tagamet over the
counter and his depression was being treated
with paxil, both of which are P450 inhibitors,
so they could operate particularly when combined
as if he had a variant in the P450 system but they were inhibiting the P450
system and that's why he wasn't able to
metabolize morphine the same way someone else would. And that's why you develop the respiratory failure. Again a genome scan
does not pick this up, an exposomic scan could have picked this up. So what are the most important
backgrounds in precision medicine? Well it's actually not genes or
environment but it's actually disease and treatment. And we need to do studies that will help physicians treat
patients if we're going to operate in the precision medicine world. Right now we're not. So why is that? Well the medical
perspective is I need help diagnosing the patient
and environmental epidemiology or exposomics and epidemiology
does help with this, risk factors are a kind of weighted variable, if you come in
with chest pain in your history of smoking and in a high
cholesterol diet, odds are you probably have a heart

But in terms of treating the patient
it's not so helpful. In fact I might ignore
the study results. The fact that I smoked or eat a high cholesterol die, I will
probably stop smoking and stop doing that and eat a much healthier diet,
but whether you do a cardiac cath or a stent or
surgery on me is completely uninfluenced by the fact that I used to smoke. The
same is true for lung cancer and smoking lung cancer
may have been caused by my smoking, and I almost certainly will stop
because now I have lung cancer, but whether or not I get surgery and which
chemotherapy I get is actually irrelevant
to whether or not I smoked in the past.

So if we're going to work in the
precision medicine world we have to right we have to answer the right questions because there are no right
answers to the wrong questions. So what are the roadblocks? Well medicine
works at the individual level. It works with respect to diagnosis,
people come to a physician if they have a problem,
they're not coming to a physician, with the exception of primary care, when they don't have a problem so there's something going on. The physician may not know what that is
yet but there is definitely a problem. So
there's a need for diagnostic help and then there's a need for treatment help
and treatment comes with side effects and treatment can have variable outcomes. But in the public health world, which is where environmental health and exposomics typically lives, we think about research at the population level. We think about risk factors, if you get exposed will you get sick, if I get
exposed to phthalates will I develop type 2 diabetes,
but it's really about probability of illness which is very different in
medical world in which certainty of illness exists.

people don't see doctors for no reason. And in environmental health and in
public health, we typically think about prevention, which is very very important,
but it's not what physicians do. And we think about susceptible populations,
uh is pregnancy a time of life in which you're more susceptible to environmental
factors. So in summary medicine really starts
after the person is sick and public health is more interested in what
happens before the person gets sick because
public health is really interested about prevention. And outside of primary care, medicine isn't all that interested
in prevention. It's not interested in why you are sick,
we're more interested in how to treat the disease that you have.

And it looks kind of like this. So public health tends to operate
in that area of time that predates the disease so
if this x-axis is time and this knot is when a disease appears,
when that is when the phenotype becomes present so on the right the phenotype is
present on the left the phenotype is not present,
public health research operates here with either prospective cohort studies
or case control studies that try to go back in time and genetic
studies that might use a family-based association design are still trying to go back in time to find
risk factors that might explain why somebody developed a disease. Whereas if you're talking about treatment, you're actually operating in
this space to the right of that notch. You're
looking at randomized controlled trials or maybe you're doing clinical
observational studies but you're trying to understand what
happens after somebody gets sick and how they respond to a treatment or a
or how their disease progresses.

So this is the area in which precision
medicine needs to operate. So most precision
medicine studies actually start after the disease begins which is not
the way most environmental health studies
operate and in precision medicine you take a group of patients and then
you move forward in time and you see how the disease progresses,
whether they're being treated or whether they're simply being observed for
their progression if it's a disease for which there's not good treatments. So in summary in public health, causation
is really the research goal, and we look at phenotypes and sub-phenotypes so
example we might look at Parkinson's and
Parkinsonian syndromes and that complicates our research because what
causes a Parkinsonian syndrome that is a disease that looks like
Parkinson's disease but actually subtly different may be different than the
causes of actual Parkinson's disease.

So sub-phenotypes may actually make it
harder to do our research. In medicine causation is less relevant. If somebody's sick and whether they have Parkinson's or
Parkinsonian disease maybe it isn't so important. Maybe if you get
exposed to lead which is neurotoxic you'll have similar side effects
to that if you already have Parkinson's or you have Parkinsonian syndrome
so maybe these sub-phenotypes that we worry about in public health research
aren't as important in the precision medicine research goal.

In public health, we want we're looking
about we're really interested in prevention
so if we're doing a case control study in particular we have to go back in time,
we want pre-clinical samples, we want pre-clinical
environment so we can understand what caused the disease. In medicine if we are to do research we don't really have to go back in time,
really just we want to have samples that are
collected after the disease starts and we want to follow people going forward. It's actually much much easier I would argue than doing public health causation
research. In public health we tend to counsel at
the population level, however healthy people
don't listen and any physician who's told a healthy patient that they need to
stop smoking can tell you that it really is difficult to get that
patient to stop smoking. On the other hand their patients that
get heart surgery or they get lung cancer,
when they're told to stop smoking they stop smoking because sick people listen
so again it's easier in some ways to do
this research in the precision medicine world because
people actually will listen to you.

Very different than the public health
world. Also if you're doing research, healthy people are harder to recruit, harder to retain, they're very very busy, there's less incentives for them to stay
involved whereas people have an illness they're
much more likely to sign up for a study and they're also more likely
to stay in that study long term so retention rates tend to be better
and in public health research we also need a control group if we're doing for
example case control study and I would argue that in precision
medicine research we actually don't need controls all we need are cohorts of
people who have a disease and we need to be able to follow them longitudinally
over time and measure their environment to see how
the environment affects the progression of the disease or the response to
treatment. So why now? Why should we be doing what
I'm going to call precision exposomics? Well
in the last 10 years, there have been exponential advances
in exposure science.

These include things like untargeted chemical assays, which
are based on the same technologies
metabolomics or high resolution metabolomics, but
we can now measure environmental chemicals as well as the metabolomes
simultaneously. We're also able to create higher
dimensional targeted chemical panels. These include both endogenous chemicals
and exogenous chemicals so we're able to actually do targeted fully quantitative
panels as well as untargeted semi-quantitative
assays. There are satellites circumnavigating
the globe that allow us to measure air pollution,
temperature, green space, light at night, and a whole host
of different environmental factors in smaller and smaller grids,
some grids as small as 30 meters square which is how temperature can actually
measure. And even things like PM 2.5, which is a
one of the major components of air pollution,
can be measured on scales of a single kilometer, whereas a few years ago they
were 10 kilometer squared scales, so we can actually measure the
environment at a much finer resolution, both temporally and geospatially than we
ever been able to do before. And then with big data we can
mine public databases, we can mine the census data and get a sense of the
socioeconomic status of where people live, we can mine medical records using natural language processing
and machine learning to actually better understand
the drugs that people take and their past history with respect to surgeries
and injuries so that we have a sense of that component of the exposome as well.

And there are wearable devices I would imagine most of the people in the
audience probably either wearing a fitbit, a garmin, or an iwatch, which is a
wearable device which tells us all kinds of information not only about
our physiologic response to the environment but where we are
geospatially. I can guarantee you that Google and Apple know where we are,
all of us right now, unless you've turned off
the tracking parts of their software, they know where we are right now, they
even know whether we're indoors or outdoors. So they're able to actually measure our environment at a scale
that is very very fine-grained geospatially and temporally. And this is all possible because of big data computational infrastructure, which
arose because of genomics over the last 10 years,
we are now capable of analyzing data on this scale. So in summary public health research is
interested in understanding vulnerable populations but I'm going to argue has
largely ignored the role of environment in response to
treatment and the effects of environment on people
with a disease, which is a subtle distinction from the
way most public health works, I'm not saying that we're not studying how
phthalates cause type 2 diabetes.

What I'm saying is
we're not taking a group of people with type 2 diabetes
and asking what happens to their insulin control
and dose if they're exposed to phthalates.
That's different. In the first example we're thinking about phthalates as a
cause of type 2 diabetes. In the second example type 2 diabetes is
the study cohort and we're saying how does
phthalates affect the treatment and that is a
different kind of design, a different kind of study,
and that is actually a precision medicine study and not a public health
study. But environmental health can operate in both
fields. So we need to start thinking about disease severity or disease
progression so people who have a disease, we can
understand how environment affects them. Because I would argue that they're
becoming the largest vulnerable population
in the country. So we think about vulnerable populations such as
people who live in poverty, people who are pregnant, children,
very very important but what about people who haven't who have illnesses
and they intersect with all those other vulnerable populations. We really need to start understanding their environments and how environment
is affecting their disease and their response to treatment.

And there's lots of ways we can do that,
we can use untargeted assays, untargeted assays may be very important in
diagnostics. Perhaps we can do case control studies
of people with bulimia versus anorexia, and we can
use untargeted assays to better understand whether or not
they have one or the other or a mix of the two because the diseases aren't
necessarily mutually exclusive but they can actually have some overlap, and allow
us to do early diagnosis or even determine the prognosis or the cause. And then if we have patients, we can follow them longitudinally
and we can look at differences in their untargeted assays and whether or not
they predict disease progression, so perhaps
there are particular signatures of these untargeted assays
that actually predict which patient with Parkinson's disease will have a highly
rapidly progressive disease versus someone who does not,
because right now it's almost impossible to predict
which patient is going to have rapidly progressive versus a stable disease
in their response to L-dopa, which is a traditional treatment for Parkinson's

So untargeted chemical assays can be run
in cohorts of patients longitudinally, and we can get a better understanding of
prognosis and response to treatment if we were to do this. We can predict risk of
complications perhaps, maybe some patients who are more likely
to get a adverse event from a drug treatment
actually can be predicted based on their own targeted
chemical screening so we can predict the response to treatment or side effects
and we can even do a lot of this work without enrolling new studies. We can
actually take archive samples from clinical cohorts or
nested within randomized controlled trials that
have already been run. Again study design matters when applying
exposomic data, we want to look at patients and we want to go
forward in time. And then there's wearable devices and the internet of
things, so as I said most of us are probably
wearing some kind of fitbit uh wearable device that exchanges data
through the internet and with other connected devices,
downloads gps and physiologic data directly,
and if you want a low-tech version of this there's a my exposome
bracelet which actually if you're wearing both a fitbit
and a my exposome bracelet you can look at both exposure and the physiologic
response to that exposure over time that was simultaneous to when you wore
the wristband.

So this wristband actually is a silicon wristband
that collects uh ambient and dermal exposures that occur that happen to
you while you're wearing it, so people typically wear it for a few days or up
to a week and it's an integrated measure of your
environment over that amount of time. So wearable devices are widely employed
in public health epidemiology, however they're not so widely employed
in disease cohorts and I think that's a place a space that we can start to move
into. I think it's starting to happen,
certainly the all of us cohort is starting to do that so I don't want to say that
no work is being done but I will say that in the environmental health world the work that's been done is relatively limited, the number of
disease cohorts that have worn uh the my exposome bracelet is
actually relatively small, but if we were to do that we can track
response to treatment, disease progression, we could look at
things like activity and exposure during chemotherapy
using fitbits and other types of wearable devices like actigraphs and if
we link them with bracelets and other types
of exposure wearables we can look at whether or not exposure
affects treatment response, we can look at things like heart rate variability
and air pollution after coronary artery bypass,
or we can link untargeted signatures to high-risk electrophysiology phenotypes
if we were to get urine or blood samples on
patients who wore fitbits.

And then there's GIS, so there's
a lot we can do. In the GIS world, we can look at air
temperature and climate, we can look at green space, we can look
at uh pollution air pollution, we can look at
uh traffic patterns and noise, access to healthy foods, SES,
uh we can even mine social media content so you can actually
look at twitter uh and there actually twitter has geospatial data in it and
this is actually a map of all the quote unquote mean tweets
in the New England area in 2015. This is actually done by
one of our faculty, who was able to get
twitter data and the geospatial locations and looked at
particular mean words and he plotted the distribution of mean words
in the tweets in New England as a measure of the social environment.

So I like punk rock, Joey Ramone talked about DDT,
one of the few rock songs that actually talked about DDT. So what can we do with exposomics in precision medicine? Well we can look
at wearable device data and how that affects patients, both in
primary care but also patients with the disease, we
can collect address history and using satellites, remember satellites
do go back in time, satellites most of the satellites that
we use for PM 2.5 went up in the air around 1999 to 2000, so
we have 20 years worth of data that we can potentially get on any given person. We can get occupational histories perhaps with exposure job exposure
matrices and if we can actually find these job
exposure matrices with untargeted assays, if we take
cohorts of people with particular types of
jobs and actually start to measure the environmental factors of them
before and after they start work.

One day I think we're going to take our
cell phone into the doctor's office and it's going to be downloaded at the
front desk and then we're going to get an untargeted chemical screen and
there's probably a genome scan that's already being logged in our
electronic medical record, and then we ping the cell phone device
which has geospatial data, which can be linked to various types of maps about
air pollution and noise and SES.

They can, AI or
artificial intelligence, in the background can generate risk scores for
the physician from the exposure in terms of their
neighborhood statistics, their occupation, chemicals they might be exposed to, air
pollution, etc. And even perhaps link that with their
genome in a genome by exposome analysis to give the physician a better
understanding of the risks of illness that a particular person has, which is a real precision medicine vision for the world.

I think it has to
move beyond genomics and start thinking about gene by environment interaction or
genome by exposome interaction. And there's a
lot of work that still has to be done in the exposome world and these are just
some examples of exposomal precision medicine projects. For example,
what if we measured phthalates in patients with cystic fibrosis to see
the severity of their lung disease so phthalates have been linked to lung
disease, such as asthma, patients with cystic fibrosis certainly
have reactive airway disease, so perhaps if a patient with cystic
fibrosis is exposed to phthalate it actually worsens the lung function.

That's a study that hasn't been done yet. We could look at environmental obesogens
as predictors of glucose control in
patients with diabetes, so either type 1 or type 2 diabetes. If
we believe that obesogens affect
glucose metabolism, then we should be studying people who already have
problems with glucose metabolism and see how these environmental exposures affect
them. Likewise we think that a lot of metals
are neurotoxic, lead, mercury, asthma excuse me arsenic,
perhaps they're modifying or exposures in the past or present or modifying
Alzheimer's disease severity and progression. We should be looking at air quality in surgical ICUs to see whether or not the
air actually has a lot of particulate matter in it
that might actually be affecting the outcomes of different surgical ICU's. We should be looking at untargeted chemical assays to detect
eating disorders such as anorexia or bulimia when they're only suspected
rather than clinically obvious. There's all kinds of potential studies,
in fact if you think about PMI or precision
medicine studies in exposomics I would argue the field is
so open that most of the ideas are really low
hanging fruit.

They're actually relatively easy to come
up with what would be a novel idea and exposomics is different um I think
you've probably figured that out by now. Exposomics actually is very different than
other omic sciences because it's not
a assay, it's not a type of assay, unlike you know the Illumina
850k chip has sort of taken over epigenomics, at least with the
methylation component of epigenomics, obviously there's other types
of epigenetic marks, but um and genomics is really about
sequencing and you know proteomics has a few
different types of measurements but they're all measuring
the same sorts of things. I think more than any other
um field exposomics is very very broad.

It will probably always include
questionnaires. We want to know where you live so we can link it to their air
pollution model. We want to know what kind of job you've
had so we can figure out what your past exposures to chemicals might have been. At the same time it is a big data field that uses GIS,
it uses geography to actually measure exposures to temperature and weather
and air pollution other types of environmental factors. Pictures could become very prominent. My daughter likes to take pictures after
she cooks a meal and send them to me so one day those pictures are going to be
used I would predict in environmental epi research or just
nutritional epi research.

Also people can take pictures of their
cleaning products in their home and that will give you a
better sense of what they're exposed to when they're cleaning their home. Mobile devices are going to play a big role, so we can actually know where you
are but also integrate that with various types of
geospatial models. When you go through a geospatial hazard
field, what's your level of exposure as of when it happened. Obviously there will
always be biomarkers such as untargeted chemical assays so will always be a need
for urine and blood samples, and then there's gonna there are big data
computational needs so that we can integrate
exposomics with the other omic fields such as epigenomics, transcriptomics,
genomics, proteomics, metabolomics, etc.

In which exposomics is actually the
top layer, or the layer in which interventions can occur, because if
you're going to affect those other omic fields you have to
start with the environment because that's actually what's driving most of
the changes in those fields over time is environment, which is largely right
now unmeasured but we're getting closer and closer to
measuring it every year. So in the future exposomics and precision
medicine will enable diagnostic tests so we can identify subpopulations with
different prognosis; we can better understand the
relationship between the chemical, nutritional, physical, and social
environment and that prognosis; we can identify optimal treatments based
on one's exposome so we can better understand response variability, which
might be due to environmental exposures or the probability of side effects or
even the probability of compliance based on how your exposome interacts with
the treatment and the disease.

And if we do this,
precision medicine programs are going to look more like this. I forget which one
this is, but this was actually the only website I was able to find on precision
medicine which wasn't completely focused on genetics but
actually talked about genetics and other individual features
such as the medical history, the environment, their behaviors
patient behaviors, habits, and biomarkers to that effect so that we can really
stratify people at an individual level because even
two identical twins are going to have very different environments and they may
respond differently to a treatment because of those different environments.
And once we do this exposomics will complete
the precision medicine puzzle. That is my last slide.
Great, thank you very much Dr.

Wright for a very fascinating talk. Just a reminder to everyone, if you have questions please submit them to the Q&A
function and you'll see that at the bottom of your screen. Also as a reminder we will be we are recording this talk and we will
put it on our website following the event. Right now I don't see any open questions
so we'll just give everyone a minute to uh to type it in. So um while the I know it usually takes
the audience uh you know a couple minutes to type in their questions
so I'll start off with a question so obviously you know from your talk uh
measuring the exposome is much more complicated than measuring the genome
right? Which is one assay.

Here you've got lots of
different things and I would say that yeah I guess our um you know our
primary care physicians are are already doing a
very simplified version of an exposo-scan every time we go see the doctor they give us a little questionnaire
you know about smoking history um other very basic things so
I was trying to think like what what do you think um
would be you know trying to add in all of this would be extremely expensive
right away right, so what do you think would be
um I guess the lowest hanging fruit, what should they be asking those
questionnaires or what um fairly inexpensive thing could
could they be measuring that may be most helpful.

Well um I don't really think it's going to be
that expensive, I don't think it's going to be any more expensive than other
precision medicine initiatives but you know obviously
we don't ask a lot and I think part of the problem
is I would argue that the average physician
has never even heard of phthalates, the average physician
wouldn't know uh what did at what to how to respond to somebody
who said that you know they drink a lot of bottled
water am I getting exposed to BPA and I think it has to start with medical
education. I think physicians really aren't
equipped to answer that. I think physicians
there's actually some research to suggest that they're
they're not equipped to understand genome scans
so you know when gives you results and you bring it to your
physician, it's almost indecipherable to him or her
and I think one of the challenges um in precision
medicine whether it's a genomic related challenge or an exposomic related
challenge is actually medical education is going
to have to change and it's going to have to adapt and I
think it doesn't mean that every physician has to be an expert in
all these fields but readily available information on
whether it's genomic or exposomic information
needs to be available and there need to be sub-specialists trained
that can deal with the more complicated issues that
they're going to arise.

So yeah it's going to have to start with
education and I think that's that precision. I think you actually had a
good example in your talk where um they they missed
it was the inability to metabolize morphine I think
and they didn't have they checked for mutation that wasn't there but they were
taking two different medications that were kind of
synergistically I guess um inhibiting the P450
right? Yes. And if the doc doctor probably would have already known
that they were taking those medications and if there was if they had just gone
through that thought process they likely could have caught that. Yeah
possibly I think over the tagamet is over the counter and
it's a P450 inhibitor and I think um think about when you talk
to your physician, do you tell them about
all the over-the-counter medicines you take?I think it.

They ask me that yes. Yeah but I think it largely gets missed but if we
measured it it could be picked up in an un it
would be picked up in an untargeted screen because pharmaceuticals
tend to be at higher dose so you're more likely to pick one up
so um that's why I actually think untargeted screens can be
very very important because it will help you
pick things up that you otherwise wouldn't pick up because it's very
difficult in a 15 minute medical visit
to ask all the relevant questions um in fact it's getting more and more
difficult as medicine becomes more commercialized and more you know
more corporate to actually be able to get all this information and information
is getting bigger and bigger and the medical visit is getting shorter and
shorter and that's a big tension that's going to
be important in precision medicine no
matter what direction it takes uh it is something that has to be
addressed at some point because uh it's just not possible to ask
all the relevant questions so that the information is going to have to be
boiled down to something relatively brief that physicians can understand.

That's good that's a good point. So we do have a few
questions typed in now so um one from Jonathan
Kim, a pretty broad question, said you gave me a lot of information
can you go over again what makes exposomics different from
epidemiology? Well genetic epidemiology is another field
I mean epidemiology is a tool in which to study
disease, what I'm talking about is more the environment just as genetics is
different than epidemiology, exposure or genomics is different epidemiology,
exposomics is one of the things um you can consider we can learn a lot from
genomics and how to approach exposomics so
one way to think of this is the difference between candidate gene
studies and genomic studies, so you miss a lot in
a candidate gene study you're really focused on genetic variants
that for which there's already information
and you're not thinking about those genetic variants for which have never
been studied before and the same thing is true of exposomics. When we ask about
smoking and we ask about high fat diets or we ask about whether
you live in a house built before 1979 which puts you at
higher risk for lead exposure, we're talking about those environmental
factors that have been well studied that we know about.

We do untargeted screens when we start to link geospatial information so that
we can see that particular clusters of diseases, a
particular cluster of exposures occur in particular geospatial
distributions, we're doing an exposomic study because we're not starting with
the limited candidates that we know about,
we're trying to discover something that is unknown and that's really what
exposomics is compared to environmental health is really about discovery, just as
genomics is different than genetics. So so we have another question from tes
mersha he asked um how do you integrate biological or omic's
data with non-biological or exposome for complex disease? Can you talk
practical experiences of integrating these types
of data? There are many variables and each may
contribute directly or indirectly or have different effect size to an

Yeah so exposomics is the newest
of the omic fields and there's a lot of work to be done. I would I often make the analogy that exposomics
is roughly where genomics was around 2005. So in 2005 we didn't have whole
genome sequencing, you know if you could measure 256 SNPs,
that was considered a genomic study. We would even use the term genomics, but
obviously that's nowhere near the genome. Right now in exposomics we can measure
thousands of chemicals but we can't measure the millions,
perhaps even hundreds of millions, of chemicals that we're exposed to,
so you know we're stuck mostly with candidate
exposure types of studies and we are also looking at technologies
that are measuring simultaneously the metabolome
and the exposome and we're making I think some mistakes
in separating them.

First of all any metabolite
was once outside your body, because it was once something that was outside your
body that you either ingested or inhaled and then you processed it so it is part
of the exposome. And we're measuring other chemicals and
how that may affect the metabolism so there's a lot of work that needs to
be done in terms of integrating non-nutritional exposomic measures
versus nutritional exposomic measures and there's a lot of work that needs to
be done in understanding how their variability
predicts variability at the epigenome level, at the transcriptome level,
and all the different omic levels so there's a lot of work that needs to be
done, but most of the variation I would argue
in transcriptomics and epigenomics is actually due to variation uh in the
exposome that we don't know how they interact and
we're still working on measuring them so the exposome is still got a long way to
go before it measures millions of chemicals
but I think it's going to get there.

There's a lot of work that's
happening that's actually sort of analogous I
would argue to microarrays so many of you may be familiar with
aptomers, which are synthetic antibodies that you can
generate to small molecules, and so you can start making customized
targeted, in theory, you can make customized targeted
microarrays for environmental measurements,
whether they're hormones that respond to stress or whether they're chemicals
that are exogenous or whether they're metabolites and you can zero in on them
much more readily than we could three or four years ago so there's a there's a
lot of things that are happening in fact the I would argue that the sort of high
resolution mass spec that we do now may not be
what we're doing in three to four years, I think there's
going to be different technologies that evolve just like genomics. You know the way we do genomics in 2005 is
radically different than the way we do it today and I would argue in 15 years
the way we do exposomics is going to be radically different but
it's not going to be one field.

I think we're going to mine
social media. I think it's good we're going to mine
satellite data and I think we're also going to do assays repeatedly over time
and then we're going to have to set up databases
that catalog all this information. Just like dbSNP catalogs the genome,
we need databases that catalog this so that researchers can find this
information, don't have to keep reinventing the wheel. Yeah thanks so I think the next question is getting at those untargeted
assays um maybe the mass spec ones that you're mentioning
he asked how do you see that the whole spectrum of environmental contaminants
like PFAS, TCE, PCBs, phthalates
microcystins and many others, how do you see them being evaluated
and their impact integrated into public health disease and healthcare? Yeah that's a great question this is
you know this is work that's starting to happen
um with toxcast and the comparative toxicogenomics database that people are
starting to catalog uh exposure to these chemicals and what
their potential health effects are and then just as genomics has gone away
from looking at single SNP disease relationships to
more complex relationships using more complex statistical approaches
you know the field of environmental health is getting
using complex mixtures and different types of data analysis
that actually integrates all these different exposures
so that we start to understand the complex mixture
that's actually occurring and it's a little more complicated because it's
time varying so in genomics at least when you're
talking about DNA sequence, the sequence is
time invariant.

So you know those SNP SNP
interactions in some ways are a little easier to understand
but because phenotypes don't appear at birth,
we're probably mismeasuring the relationship between
these sort of complex SNP-SNP interactions and disease
um because of that. In environment you know we're working in
the same sort of area looking at the complex mixtures
and a lot of works actually happen a lot of work is happening at michigan
actually in the statistical area of complex mixtures so
exposomics is really about measuring the environment but there is a whole
field of informatics and statistics which is growing in
parallel to try to take that data and integrate
it with other omics and understand its relationship to
health, so both of those things have to happen

So the the next question, you're very popular
with questions today, uh the next question what is the
implication if we find for example phthalates
influence glucose metabolism in diabetes patients? So you know if we find an um influence like this from the exposome do we educate physicians to be aware of the phthalates or at the
population level do we try to lower phthalates exposure or like us how do we
react? Well great question well one of the reasons why I'm a big
proponent for studying environment in patients is because
if you give a physician something they can do for a patient,
you get their attention. If you're talking about
a risk for somebody getting sick, they may bring it up,
the person may not listen, this is sort of how smoking and heart disease
works at the physician patient level.

If on the other hand
you have a patient with diabetes who their insulin control is very poor and
you know they keep increasing the dose and they can't understand what's going
on and it turns out they use a lot of perfumes, perfumes have
phthalates so they're one of what perhaps one of the reasons
why their insulin control is so bad is because they're using a lot of perfumes
and the physician can say well let's try not using the perfumes because they have
a lot of phthalates in them and see how that affects
your insulin and glucose control and it works
then all of a sudden medical education is going to change. So I think if we start targeting patients with some of the information we
already know about how these chemicals work
then we can actually sort of subtly insert ourselves into medical education
because once there's an actionable intervention
that a physician can do for a patient, it changes the dynamic enormously.

And I think if we can show that there's something we can do,
which we can often do with exposomics, and remember patients are
more sensitive because they have a disease so the effects are probably a
lot bigger than we're anticipating because
we're so used to studying healthy people we see these tiny effects. When you study someone who is already sick, you may see a very big effect. It
may shock you how big that effect is, because they're already sensitive and
vulnerable and I think that's when medical education is going to change is
when we can do something for patients.

But yeah I think that's a really
important thing going forward. Yeah thanks that's that's great um the
next it will do two more questions and we'll
wrap up. Next one is so you gave a lot of
examples of channels to data mine information for
exposomics, fitbit, social media, maybe even grocery
purchasing patterns, what sort of ethic questions or
challenges do you see arising from this information? Should we be moving toward
designating these things as private medical information
and therefore protected under HIPAA
and what are the ramifications of that? Oh a lot of them are
um any geospatial analysis requires understanding where you live and that is
that's considered PHI, personal health information, so a lot of
it already is.

I think as a society we have to start
thinking and maybe having a discussion over
the value of analyzing this data for health reasons versus
commercial reasons. So I can tell you that
Google, for reasons I'll never understand, um
Google and Apple know where you are right now,
and they don't have to get consent. You just have to agree to download their
software. And if you have an iPhone the health
app actually has a way to download your medical record
with just a single click and so your medical record will be downloaded
onto Apple's cloud, so Apple will have your medical records
and they did not get consent. I'm not arguing that we should go in that
direction, in fact I think we need to get consent,
what I'm concerned about is that there's a double standard
where researchers are not trusted with the data
that commercial entities like apple and google
are, and perhaps we should have a public debate over whether or not that's
appropriate because Apple and Google I guarantee you
are using that data for commercial purposes
and so I think there is a lot of problems with the way we handle
health information in the public realm where companies
don't have to get consent to do basically what researchers are trying to

And I think there needs to be a debate
about that because I would prefer we get consent and I
think we should get consent if we're going to
understand you know the implications of health but we're
doing it overtly with consent and Apple and Google are doing it in
much bigger numbers of people because they don't have to get
consent. I'll be honest if I was a young data scientist, I would have access to much more interesting data if I worked for Apple
or Google than if I worked for you know a university,
because Apple and Google have enormous amounts of that data
and we have to go through a lot of processes and regulations to get that
data and there is something wrong about that.

And I don't believe it's wrong that we have to go through the process, I
think it's wrong that Apple Google don't have to go through the process. Yeah some interesting points there. Okay last question from uh it's from
Ruby Hickman she asks in informing treatment, would
you be looking at are trying to find a small number of
height effect exposures or how would measures like untargeted
assays inform treatment choice or dosage? Well I think it's an unknown how big the
effects are going to be because there's almost no research in the area. My gut belief is there's going to be some really big effects that we were
just unaware of because we never studied them
and I think they might be in areas like type 2 diabetes or type 1 diabetes
or or more severe diseases.

There is there I don't want to say
there's been no work in this, certainly the whole field of asthma your asthma is
very much an environmental disease, there's a lot of work
on how environment affects asthma, but you leave that
disease and there's very little work. A really great study which was published
out from the University of Washington a couple years ago
actually showed that patients who had lung transplants
who lived in areas of lower air pollution had longer
five-year survival rates than patients who live in high air pollution areas
and there's some there's some really interesting work that they
can be done that's going to show relatively big effects
and I think we don't know what they are yet because this is
really a field that environmental health has not been particularly interested in
in the past but if we're going to get into the precision medicine world we
have to get interested in it and I suspect
there are some big effects.

There's probably a lot of small effects, I think
you're right about that, but I also think because this is largely
uncharted territory there are probably big effects as well. Thank you, Dr. Wright. I think that's all
the time we have for today. We appreciate you taking all of those questions
and I'd just like to say to everyone who's still with us
thank you again for joining us for this wonderful webinar
as part of our summer omics learning series and
uh for those of you who will be joining us for the discussion
uh please see your email from me earlier today. That will have instructions for joining and if you have any questions
please just let me know via email.

Thank you and have a great day. Thanks



Please enter your comment!
Please enter your name here