Embed code
Note: this content has been automatically generated.
Uh after finding out last night many of
you from or nutrition background. And
less of the market by background to
explain a little bit of the history
have some of the inscrutable diagrams
saying at the last year as as I I've
lost a site. And also some of the how
we developed Some of the techniques
lead to those so begin with that then
talk about how we apply them problems
of AB this remote users so the
backgrounds well the verses as you IDNA
sequencing getting dramatically cheaper
so much so that would be quite
sequences energy lyman two thousand
nine it was literally almost a million
times cheaper than the for the beam
into thousand one when the human genome
project was announced complete and it's
got it's got a couple bottles to make
the CT something. But then the problem
is you have to interpret all of that
sequence data so what I'm saying here
is just nine out of a hundred thirty
thousand sequences project we did with
the first lap about the reason like
this is what happens if you use a file
containing the final genetic tree the
related sequences to each other. And
that's what happens if you just
visualise the tree. So you can see a me
really that you have your work cut out
for you interpreting all the sequence a
local street data. And that doesn't
even help me that much I tell you study
design for this particular project what
we were doing this by geography but not
on the grand scale and one Wallace have
the nineteenth century spores you match
the distribution of logical organisms
across the we was study my card so we
thought would be able to find by
geographical patents much places time.
So what we did was we took three
keyboards and then I think it's it's of
the people we typed a small all the
keys not think is hands and initially
we just wanted to answer basic
questions by geography like the
spacebar have more markets are not
simply because a lot of the mail but
he's that is that is that with you my
could survive it lush valleys your the
prince. And so it was here by
geographical dividing line between say
the GQBHP because the the right the
right next to each other on the
keyboards as you type in there of any
type of opposite hands. So depending on
how the space to what you might see
different patterns. So that the the
data registry idea makes the main thing
going on on the system immediately
clear so on this diagram each point
represents. all the markers and on time
I heard look you know she identified by
the DNA And so it's points comes from
either negative or from a key on the
keyboard defined colour them by paced
and you see the points this person of
the second person and that that this a
cluster together or some contrast if I
call them according to what he but I
think it you see my separation. So you
can tell immediately that we each have
a unique personal skin community we
transferred from I think it's it's like
people to suite at another schematic
PNES few LC but more importantly labels
on the TVCCSI Miami so you really know
it straight so so fine that aside so
also I'll tell you a little better
that's we produce these diagrams what
was also principal coordinates plus you
different distances well explain what
lays me later I one one one key problem
with be applying to you as a beastly
which is a math problem globally not
just in the united states but
increasingly in the rest of the world.
And because there are many different
actively contribute to to ABC
everything from psychology to die and
so fourth. And there are many human
genes that contribute to obesity this
is just the buttons just a recent
review assuming some of the ACL say
documents that we now know contribute
to obesity. But the problem is that the
project and the value of these generic
models a very porous so if you take
every humans that ever identified by D
bosses contributing to it easy you can
only predict easily and he's a piece
with about fifty eight detectors which
is not very impressive say says a lot
of work that I'm gonna sell you balance
in collaboration with yeah and about
about a decade ago he became one of the
leading I don't suppose hypothesis that
the obesity epidemic might actually be
an epidemic in terms of the transmits
possible agent. And the first what's
really but between the these
individuals to different the market all
communities you've already seen some
data from disk a and from Frederick
other person straight but yesterday we
would expect change weights to Carly
changes microbial communities and we
will also explain altering the
microbial communities should also
weight gain they lost. Um and say what
what I'm gonna show you here building
on what you saw yesterday is that all
three of these two humans sorry all
three of these attribute mice buffet
straight in the face to intrude humans
amalgamated helen's at the moment is to
figure out how we do that. and humans
as well. Um solicitous of this problem
of how we measure at least microbial
communities which are very calm place
again as you ideas today there millions
of jeans and we got hundreds of
difference P C.'s and you can just slap
annotate major figure out you can you
like. Um and in this context maybe but
still thinking be colours a classic got
my great but the reason why we think
that is not the so that is already to
the entire again me and again pretty a
bacterial in the file and as a major
player and they got microbe I'm it's
just that you call I was really get a
living in captivity unlike the vast
majority of my kids that are out there
in the environment and that are in
there Larry got and say figuring out I
have to deal with this and culture
majority involves a lot of the way you
hold that that you had about yesterday
and a lot but you will hear about
today. Um say so so you a lot of one
and gives tell you about today is gonna
be based on studies of the rubber
similarity G other tell you about that
X number infantry which pieces of
presents we got I I believed in long is
get a tell you much more about change
management mix we yes where where you
get a view of the functions well and
what I'm getting like here is looking
at this P C.'s which is a very cheap
way of doing these studies can tell you
a lot about about what you want to know
about different systems although it's
only one piece of the puzzle and
getting of the functions but much
recommended genetic sequencing event by
by the transcript I mix by the tablets
and so one is gonna be essential for
police station make innocence behind
many of these patents and show you But
the reason why we use rubber similar in
a is general P C.'s as part of the
writers and so also I hazard it has
fast evolving sly revolving regions so
we can we can truck phylogeny
relationships are different dates and
then there are these huge preexisting
databases we can Paris and so we don't
have to do everything from scratch from
scratch every time we can look out a
lot of us pieces that sample in these
existing databases so what we do direct
environmental sequencing to see the
vast majority of my cards will not be
able to go to the lab so the basic
principle here is to get the samples
extract the DNA so maybe maybe the to
comes from someone lean well we from
someone is that these in what I'm gonna
show you it's mostly based on PCR
amplification small subunit runs them
or lady of of course you can take all
the DNAD shotgun sequencing out as well
which you'll hear more about ways for
those what you also good bass several
I'm just today traditionally you take
the DNAG weak cleaners and you would
sequencers but the great thing about
agency. and saying as you can sequence
direct from the PCR product say every
time you run the high C we collect
about another but in sequences. So
we're saving a billions we expect
colonies abandon including files and a
fair amount of expense and there Um and
the traditional you take the sequences
and blast immigrate them according to
what was the motive them and inbound
but ironically it's a success this
culture in the you know methods that
makes at least less useful because now
instead of having and instead of having
be like them or something else what the
name of the phenotype you recognise you
get ahead and cultured environmental I
slept number thirty eight minutes seven
one of these studies. So what we have
to do it is taken explicitly pilot
genetic approach where you do multiple
sequence alignment bodybuilder tree and
then you relational sequences you found
that our genetic tree to the
environmental samples looked at but
that she based approach has its own
problems so this picture is a pretty
hard to understand the to analyse what
I'm saying here is data from we did
with the ones that back in two thousand
five looking at five thousand sequences
that the mouse got and twelve thousand
sequences a human column from their
problems like in red and blue ones
respectively and at the time this was a
huge number of C concerns about now
surrounding your for sample what you
get back from the sequence but what you
can see there's a bunch of red clusters
in the budget clusters at the time we
did this there was significant
statistically tells the reasonably
distributions will not identical to
each other. And that might not seem so
that on the street what what two
colours but I haven't ODS as that
actually we were looking at twenty one
different even samples and samples from
nineteen different mice but that I we
paper battery forty different colours
you try to figure out a sigh and most
similar so a well also a three looking
at a buyer I you really have your work
cut out and just illustrate the problem
the goal of this is not a company's
particulars as just the paper that was
at the top of my stack when I was
together what we're doing is very
valuable right there looking new right
we got out him I god they could be
involved in like this endless
degradation by feels about the issue is
you Reading through the paper you come
to think it through it's pilot genetic
tree you sequences safe and had this
little short labels the longer
sequences in the database of these
longer labels fixation numbers and so
you can say whether new sequences
better that's a very well. But then you
come to figure for which is another
tree forget five and think a six and a
whole lot of additional thing is what
the tragic and say what we had to do
was we had to get out of that first of
individual is inside capitalise a pain
really that's even to my lap you know
has an independent faculty position of
Denver the the put together this really
nice H decreases based on the side of
calculating a community distance metric
based on evolutionary history. And so
the idea is very simple so the idea is
that if you have unrelated communities
all of the branches on the tree again
to leave only to members of the red
community or only yeah the bleep
community but not about right because
you can have something from one
community moving to the other and
surviving reproducing in continuing to
add gradually. So that communities
around releases we define that just
there's one there's far out as they can
possibly be wasn't contrast if we have
two identical communities with complete
something every my could you find of
the red community you find blue. So all
the branch think much reshaped by by
communities should your paypal and we
define that distances era. And then if
you have two related communities just
see some shared branches paypal and
some unique ones available we we define
the unique fraction or you know
practise distance on the tree as a
fraction of the tree that leads at
least one community rubber leading to
but so what's so great about what's
well what we can do is we can use that
one's inside that there is one tree of
life makes all cousins to take my could
you any environment item on the same
tree so here we have one tree with the
red yellow and blue microbial
communities coloured. And say we can
compare using that unified technique I
just showed you red yellow relatively
relatively summarise all of these
distances in the distance matrix. And
then once we have that distance matrix
we can use techniques I principal
components analysis hierarchical
clustering another mo. right
statistical she needs to relate all the
communities one another at the whole
community level using the evolutionary
history the organisms that are involved
Um can spoke of is relatively similar
components analysis which many of you
have probably seen on the directory
context the differences instead of
starting with the actual what what the
actual locations and euclidean space
example what we start button's
distances no we construct a set of
principles that's all the distance
constraints and then we do
dimensionality reduction et cetera
points. So it's essentially like having
a having a flashlight where your
prediction shadows on the wall of the
points you data cloud really want to
raise a bachelor right right a that
fishing spare data points as much as
possible to explain as much of the
variability that strip this principle
access a menu right spaces at ninety
degrees to match to spread out points
as much as possible again that's just
taken principal axes and then you keep
on doing Nash explaining less and less
of the variation with each additional
access seven bicycle parts you saw
yesterday and that you'll see today
based on this technique of the success
of axes are explaining as much as
possible of the variability in the data
say other arbitrary coordinates you
derive intrinsically from the data
itself. So what does technique and and
camping initially decided she would
like us everything and say we went and
then put together this matter analysis
of a hundred studies two hundred
samples twenty thousand sequences
literally spending the will to
everything supposed to be there seem to
temperature PH just about every
imaginable fact to we could affect
microbial communities globally. And so
the question is out of all this
complexity just like all the complexity
that relates different people would
they be any clear patents that made
from the data. And much to my surprise
the answer was yes when you take all
these differences and always
environments you see this very this
last between the salient and than on
sale in environments use imagine a
hierarchical clustering as well as the
peak away this let's say station but
displays you'll less analysis piece
with live in the middle there what
those are mysteries where you truly
high. the mixture between the salient
but on sale in so we can catch of these
biological relationships among these
very heterogeneous environments. Um but
the cool thing about sort of really
matters how you measure these distances
you could people referring to do some
different distance matrix yesterday the
reason why you want to take evolution
into account when you mention based
distance Um this is the the slightly
updated you have that same data what
salient unsettling split the mysteries
in the matter with you know frank if
instead you take it able to access like
it's P C.'s level the genus level the
family level or whatever and you
customers data use of the distance
matrix like euclidean distances which
is very commonly used focus on
sequenced its metric measure or
whatever you get this spike artifacts
by two degrees relative. one another
because the data matrices very sparse
And you release things like the
biological significance of the S trees
begin to mediate as opposed to going
off in the right direction somewhere
and so the reason why is that when you
use a file genetically based distance
major you take the evolutionary history
into account when you don't do that
it's not you make no assumptions about
the rate you're actually you're
actually assuming that reassessed a
phylogeny where all be Caesar equally
related to one another and the matter
whatever seem may compile genetic
reconstruction whatever apology you
build is gonna be based on them at stop
phylogeny and explaining the biological
relationship and so you get
biologically more meaningful clustering
out anyway what one thing that
suppresses about this environmental
data it's not related to hasten to
humans just maybe one one thing that
the one thing that surprised us about
it was that extreme environments are
from our perspective without wires
wiper microbial perspective. So if you
high temperature environments like for
marine sediments from hot Springs yeah
we stayed but it's just like like the
other salient non so samples risk
actively about my colleague an
additional samples if ones that we're
able to find some really extreme
environments microbial perspective the
when we add them to the graph
introduces new access that explains
twice as much of the initial so silly
nonsense but I said it up. And and and
squash all these points from all over
the well just happen to one side was
graph. So you might be wondering how I
you have to get find those microbial
use extreme environments the answers
that you don't have to get started all
the right there with honest. say what
we're saying is mostly the million gas
as a clear our wireless part other
heist associated habitat site map
pictures here vagina not pictured here
and so forth are intermediate between
the two and then these these has to say
she's environments very different from
what we find free living or what we
find associated with them better
precise like sponges or are addressed
real better but So you save mice about
actually sector's been based on
traditional saying a sequencing and
what really allowed shifts to hire
people techniques with the observation
that we could just take a small
fragment the sixteen it G get the same
community clustering information as
what polling sequence reason why this
is really important is the the short we
technologies a much cheaper P sequence
other traditional thing the sequencing
all right now they're starting to catch
up with the reflectance anger. So this
is less but what we did was basically
took three datasets we we had me
appalling sequences from sixteen
described similar in a that's what
would happen if we only had a hundred
to two hundred fifty basis that
sequence but we get the same result or
different so what I'm saying here's
clustering of the humans and the bias
but I said you before elementary and
then they go raring agree microbial mat
which is a hyper silly green mask
that's the most this microbial
community we know about it and you can
say that despite that that this T the
humans and the buys a pretty different
from one another. So it's cut a long
story short if you take a good two
hundred base fragment you get exactly
the same clustering path right down to
the same as you can also take a very
bad two hundred base fragment from the
same they and in that case you would
make the amazing. every but some of the
humans almost similar to the mice but
they ask the other humans. So those
would be very exciting faced try you
get back to clinical data that's what
is it about these people that make
resemble the mice and have a vegetarian
till eleven restricted environment a
very tiny a well and then and then So
when you got a fully sequences and you
found out just wasn't sure you might be
disappointed but we get is also sixteen
assistant changing very rapidly
ascending uniquely tied substitutions
accumulating every minute noodles I and
so the eh regions are we getting two
thousand seven when we did this and the
same regions really get today and you
should just look at the region that's
the right wing for the technology you
use and so does it say my kind of the
this is no we can address by recap or a
say is now we met in maybe and number
of bubbles. in my lap but together the
software pipeline coach I'm sense a
qualitative insights and Mike really
colour Um it's free it's open source
run anywhere from the laptop to the
Amazon DCT class. So the hot CP
computer escort hundred forty thousand
calls just larger social run on to date
well I see there's alleged integrate
the analysis of hundreds of samples
simultaneously mostly we use that for
sixteen S and become sequencing you can
also use it for other marketers and you
can also use the shotgun method to
mimic scimitar transcript I'm it's now
almost right although those protocols
still experimental stage. So basically
what we do is we tag hundreds of
samples of DNA pockets we introduce we
make some together and initially we
sequence one four five four well they
more recently would switch to the
aluminium platform which answer so you
have to consider buttons maybe because
a sequence of the bucket we can tell
what sample came from so we could play
stop the multiple sequence alignment
build phylogeny is and the use of
powered least cost the data builders
principal K whatnot spots but you'll
see many of already about our shows you
can say you chase more say incidentally
closely coming from a nutrition
background you might be wondering vices
technology called for five for it ends
up that Fahrenheit for fifty four as
the temperature your money bins as soon
as you get into sequencing. So that's
getting cheaper all the time. But it's
still not free and we have to do a lot
of action every batch however have able
but face time we did this in the in the
two thousand save them so we we spent
twelve thousand dollars on a four five
four run and got back and and and got
back about a half a million sequences.
And set at the time of silence of the
second facility down the whole my lap
was still charging eight bucks to read.
So if we double there incident with the
new technology what it cost four
million dollars and still twelve
thousand and they wouldn't be finished
with the sequencing yes and what the
aluminium technologies stopped another
couple of orders of magnitude cost
since then say services listens
tremendous in terms of making more
research questions more accessible you
can literally forty thousand samples
now and it's not a big deal. Um say
celeste cost LA no no censors actors
lab but he's not here Nestle and well I
put together this initial map of the
human microbe I'm back in two thousand
nine and said but how do you market
combine studies up to this point is
people studying different parts of the
marker by might them out this skin this
is still that use different techniques
on different subjects on different body
side and figuring out what was easy
different people what was use different
body size what was used different
technologies was really hard. So what
we did is we rounded up run that seven
people and swap them out a swap them a
lot up to twenty seven locations on the
body which as you can imagine as a lot
of places to stick a teacher mistake in
the morning and what we see emerging
from is very clear path where the mouse
is very separate we got three separate
the skin as a lot more spread out as
their the habitat five nostrils
previously explored primarily by by was
the year and the here and so forth. And
so I give the same they said a colour
by different no different different
variable. you can say that sex has very
little effect well estimate is has very
little effect compared to the body side
with one it's So what about somebody
set up on the day something said by the
way I was on the antibiotics separable
a couple weeks ago for sinus infection
do you think that a matter I said most
late today Gary ethics committees we
could someone else was point so I
suppose we'll find out but this was
totally different not just got but also
skin and the out for everyone no study
of like three months later I know this
is what everyone else. So the study
before I was to look at everyone waited
a look at them again then wait three
months like at the beginning that way
today look again. So we could look at
different temporal scales as well. And
so what's cool about this stuff to give
us based overall picture variability
making them combine. I I don't expect
you to read the text on labels and
probably don't really care I just wanna
say that's a lot of left right symmetry
in the major Texas across the body. And
then here you you know frank distance
on the Y axis so a lot to bob people
are more different eighty samples from
the same body have habitats a bite them
out a less different two different
habitats button a habitat to from the
same this less different different
different people within a habitat
person to like today and pass a list
different collected three months apart
them and that was my staff site we
looked at the skin of the scantily
stable and then the guy was in that way
but else was kind of frustrating
because you really want to figure out
dynamics right so now study we liked it
for time points purpose and and the
human microbiology we spent a hundred
seventy three million dollars been ages
money looking at a maximum of three
time points papers. And the reason why
this is a problem so this is the
viewing software that we developed a
workable stuff so this a god with them
out and so on and if I wrote it as
around to give you an explicit access
the time you can you can see the three
month separation you can't really see
the one day separation on the scale the
kind of looks like some changes but
it's kind of but it's hard to quantify
the state is what you'd really like to
do is you really like to get back to
some of the same people in sequence and
every single day for like a year and a
half. So what the after they wouldn't
platform we could afford to do that and
it looks like this. And the concordance
between the two sequencing technology
is really nice right so you get very
comparable results when you do it
across different platforms if you use
the right it needs to collect the data.
And the great thing about that so that
we can actually starts to animation so
what I'm gonna show you is a six once
in the last two people the documents in
amoeba like points my have remanded
looking skin them out in this so if I
just I was scary you can immediately
see yourself just how terrible. yeah
there's I have constant the mountains
and have a gas into meeting And you can
just imagine have a sort of high
resolution tracking could be amazing
looking both healthy variation which is
what we see here and also looking at
people's responses to different
individuals whether the drug
interventions or nutritional or what
are the other thing that's interesting
here's what by the time we were
together and have all kinds of
opportunities to exchange markets with
one another you can see that we really
maintain as separate microbial identity
statistics one time that's mostly
around us so you can see a separation
but we have to market items I propose
period. So each of us has as individual
market by that's relatively stable at
the time okay so sit site and that's
what we disappointed where our markets
come problem the first place. And if
you have dogs orchid society you
probably have some dark suspicions
about that all but well which turned
out to be true by the way see just like
we can match you up to mouse whatever
ninety five percent accuracy we can
actually up to your dog by the market
easier. Um but all seriousness once
about what what happens immediately
after the most of what we did with
Maria accurate to make the smell like
an out in my you at at at at the
delivery might has a huge effect on the
initial. michael's so I think about the
regular way basically markets all of
your body what like models vegetable
community specifically worked in
contrast to deliver by C section all
your markets like like skin yes another
view that same day so with the bachelor
samples from the from all the models
and read all the body have chance to
actually do with the babies and paying.
And I hear all of the all of the skin
from the models and and that we all the
body habitat small see section babies
Michael And we think that this initial
inoculation might explain some of the
differences between batch release
deliver the see section babies that
again you cared about yesterday so I
would reiterate now we also look at how
the marker by undeveloped infants of
those reply looking at one of the one
kid I've that I'm starting with the
make Acadia and then getting at the two
and a half years supply so you have
what looks like a relatively smooth
progression ultimately resembling the
mobile sample what the steady increase
and I basically and service progression
looks pretty smooth must be better so
you another view of the later
integration with human market by
project data that might lead easy think
a little differently about it. And we
can also do this sort of thing cross
culturally so what it's called lab we
looked at we wanted for people in three
populations from so what I'm saying
here is just the kids is zero right to
eighteen years of why and for each
point we're looking at a distance
between that point the average of
adults most community. So what you see
as as profound change over the face
three years the black followed like
convergence more or less the adult
community by the time you have H three
analysis replicator an african
population read stuff American
population green then the US population
and blue. very more time courses.
However way you wind up depends very
much more population you Um so if you
do a principal components part of the
same data puzzles coloured by age for
the young kids out of the adults and
like the different populations you can
see the US population as profoundly
different from allowing the under and
the amber indians. And of the bible and
we don't know how much of this is due
to die how much to genetics how much
environmental exposure because all of
these things very in different
populations. And that's context and
it's important to remember the project
slightly to they and also like ms
european cable and other very important
but there and exploring this much like
this to be looking at looking at least
an adult and the same as you go to
other populations and the scene is you
going to kids you find all these I
provide configurations you just don't
see matt initial population and that's
really really important for for the
drawing conclusions whether it's about
drugs or whether it's about nutrition
in one population and then trying to
apply these conclusions to a completely
different population what the
completely different marker by now that
we know that might provide has all
these effects on nutrition that you had
about yesterday and the you'll so hear
about today so was that have of the
rest of the database to buy a bike
happening with that they all American
got up with the idea was that we could
use the drop and sequencing tighten up
the ability to participate in this
research to anybody who was interested
and so you wouldn't pay a hundred
million dollars to find out what was
and you got with the drop cost and
sequencing technology we can do it for
a donation of about a hundred dollars.
So that's point will probably raised
about a million dollars for members of
the public almost all about ninety nine
dollar increments for people interested
in putting themselves lament microbial
mat and say and services and a Mcdonald
handing out one of the active biology
student smile at a hunting acted so the
person would you in project meeting
last year where are we have four
hundred people you have an if humans
you don't see as well as having them I
combine sequence so we can look at
relationships between them and we we
have a paper with replaying to inspect
a command excel in a couple weeks not
almost population but on a related what
looking at some of the terrible
components of the market by the
demographics lettuce cover a lot more
range than previous studies say upset
for example relative to the HMP we have
a lot more older people and we have
younger people we also have a lot more
obese people about the way people who
were primarily excluded from the HMP on
the grounds of their health conditions.
Um but the overall result we get a very
consistent what what the study's so
looking at the distribution of the P
C.s the the are samples of skin samples
comparing what we if self collected for
members of the public male than room
temperature this is what was clinically
collective immediately my case analysis
which yes and talk a little about how
the market by relates to genetics and
health. And one thing that was very
surprising to us services wait what
this lab and two thousand nine looking
one is a got it this is dies I collect
twins the relationship with a basically
so when we like them on as a got twins
on average the community's not
statistically significantly more
similar up and dies I got twins and
then and then looking S what
relationship to family members of more
similar including the files as we
should not more study but no yeah it's
an ink about all twenty twelve one and
then the family study the dogs I told
you about but unrelated it does appear
there away. So on the face but also
tell you that was no you genetic
effects on the microbe I'm but that's
very difficult to reconcile with what
we know but mice so for example in the
the and and the the the what uses which
which rhetoric back it was one of the
collaborators novels slip by with laid
out "'cause" now I see this profound
alteration of the market by in the
eyepiece mice oh but then you have this
problem with establishing causality you
don't have to take community as course
or result we at least. Um then what
what then with his group at Emory look
to the different system that's yellow
five knockout mice. So again they
become obese relative to wild type
mouse I again this profoundly also
microbial community between the T laugh
part knockouts about type correlated
with high I had as their track that's
right in cholesterol and even high
blood pressure. And test originally
fascination lee what's going on here
you can even transmit this ABC to
another mouse by transmitting the also
microbial communities so you have a
mutation that triggers this change
microbial community but then when you
transmit that microbial community to
genetically normal mouse is being
raised in free but now I'm what they
market outside that mouse also become a
beast the fascinating thing with the
system is the reason why they become a
base. So it turns out the limelight
BIBIB model where it seems like that's
the difference energy have us but the
elephant knockouts what you're saying
is but what what you're seeing is
primarily a difference in behaviour
somebody a five block at each more than
about type mouse that's I did become a
beast and similarly the gym pretty my
so the genetically normal again was
also microbial community also become
hungrier also libraries and also become
a these so it's a behavioural
phenotype. We had no idea how this
works I used was impossible mechanisms
and in in in Frederick stroke yesterday
but another that and we have the
scouting drawn just remind you that you
got my critics massively outnumber your
own cells in your body maybe a poison
you when it comes time to make choices
cafeteria one thing that's interesting
is the the mice diet has a much larger
effect and you decide so what the
decreases great last year what we did
was we put fifty two back strains of
mice on high that that simply this is
like that that's great you can save the
directory fate is much larger than the
genus activate. Um and you might be
wondering if any of this since humans
so what the with the the winter wasn't
yes that what they do they put people
on a die for year what resource changes
microbial community that led them to
resemble Colleen community and that
regularly the more individual lost the
big at the change in the microbial
community. Um we we are ten nights it
was a very talented that's to my lab is
yeah I remember the administrator was
also able to shave in that in the next
two thousand nine at dataset other we
can classify people we were peaceful
ninety percent accuracy based on only
got combine and this goes also showed
most certainly in the micro be
submission populations but you get very
good classifier accuracy for at least
now that's why not be very impressive
to the commercial odious previously you
this probably easier ways to tell if
someone's obese right and sequencing
other migrates but on the other hand
remember we can only do this with fifty
eight percent accuracy based on human
genes this is ninety percent accuracy
based on the market by on that and then
and then also in humans what we see is
a long tail but not sure to entire
correlates with major several patents
the market combine especially the mouse
a crap as hell and back to reality said
this is what what care it was great
okay however I however we needed
intervention study of the short too
what we find is that individuals
clusters so whether we put these it's a
people points colours coloured by
individual here these are people who
worked in place of a low fat diet or
hike at that what you can see is all
the points the same person class
together. And say are clearly longterm
dietary changes that be needed to make
these large changes in my by one other
thing I should point out is that by
mark as are problematic widespread
basically I end for diabetes level
studies have been done across nations
we have a paper that just came out its
little services really hot off the
press china for IBD you have a
consistent signal across different
populations by liability in terms of
the markets are involved we did not see
that in the different that what sort
been studied for at least three say a
sizable there was somehow what's best
as difference in the back tried it is
to immediate ratio that is not
generally tree and we need to look for
more subtle by Marcus probably high
level ecological patents relation
different people. So you might be
wondering can we do anything about that
actually changed my combine to of
disease but we going what this is
trying to develop kinda this last made
some based not on very subtle
differences in the human genome based
on much more obvious differences in the
market by "'em" a we can deploy even
and about in countries it's a tiny have
to K then then you guys to decide out
second you know like the spider
malnutrition study site what what the
Malawi negates project where and and
the you can see for example that
despite the fact that we discussed
population they are able to afford cell
phones. And what's fascinating about
the sincere now I have a five billion
active cell phones on it which is
pretty impressive when you consider the
only seven but people up the balloon
force people in the world twenty
percent of them have their itself and
and the one the reason for this most
floor it's a cheap set up digital
signal processing because of the
decrease in computation cost but it's
effectively break but stay not as to
allowing that for example I take the
spate of dark a couple of years ago
where right next to being able to buy
fairly sizeable thanks of become you
can get yourself minutes recharge and
just a couple months yeah I was in
Tanzania what the how to hunt together
is the last time together as in east
Africa a couple of them have cell
phones late raises a serious of a train
each fill cell phones to budget I get
an extra flavour a make a hundred for
additional cell phone minutes but one
thing was getting cheaper even faster
than computation which is an Bibles why
why is DNA sequencing sometime the
computation dropping cost by a factor
of a hundred DNA sequencing drafting
cost by almost a factor Edmonton and so
the question is can we I can we
designed essays based primarily on DNA
sequencing also safety but it's free
and that's robust enough that we can
deploy these these developing countries
to really do something about some of
these major problems. Um so what else I
would be doing these studies in human
eyes last runner and I'd projects maybe
stay and our are gets funded projects
our nutrition we take samples from
individual humans of them into my eyes
and then see how much weight space mice
gain all these depending on what this
and my combine the got and the results
we get the message you know for
services also supposes time this is
principal kind of one of the microbial
community with five my colonised the
same data used became Nassau switched
them why we I might same content
seaweed see the ship censorship to the
micro by you see recovery what say you
know database I you see if it's used in
the clinic but then you switch them
back to them why would I may cry again
you can model each of these periods
just one single exponential decay curve
so we should make couple paper last
year that we can transfer individual
phenotype from people in mice again
this is like what yes that say I case
of caustic already transcribe cost my
combine from from a kid with
malnutrition and mice the musty very
badly they lease thirty percent body
mass with them there's three weeks you
leave the country they die you can
basically the same you enough also the
supplement is clinically and you can
say versus couple to he change mica by
I however more people now suffer from
PC the malnutrition and so you might be
wondering if we can do this for a
beastly as well. Uh just summarises
briefly essentially the same type
transplants also work for at least the
so you can see there's a lean databases
maybe stated which is a very stable and
different colonisation the mice
functional stability these transplants
as a one so you can see essentially
paper correlations between the input in
the at the community in terms of their
PC I'm also a second marriage and we a
function. Um then the body composition
that master in a whether you have we or
like beast either. Um we can correlate
changes in particular tax or what
particular changes in the tabloids
would have let me clarify I would I
would really talk about that to mention
that we see it changes in sorting fatty
acid metabolism example as you might
expect depending on with the with the
tighter human this wiener index. Um and
then when we look across more vehemence
level so looking at a transcripts
metabolites and and number sixteen is
profiles are essentially what you see
as agreements these different levels of
analysis. So we think we might even be
able to predict the metabolite from
eighty nine vice versa with the
metadata. Um and well I was really cool
so we can use culture collections where
we take hundreds of strains we culture
out of single individual and
reconstitute Armand and whether we
using the un coded community what what
highly sorry what's all the points here
all the coaching community with holly
points after the release also balloon
data we get a very similar community
forms that individualised and then the
correlation on the functions is
extremely is extremely good will be
using and coach unity all code.
communities them but and so what's
fascinating about this is we can
finally taste ecological hypotheses
about what's important in getting
causality by leaving out the market we
think a really important. And then
bringing in the specific strains we
think again to have a lot to take
computer type So there's a character as
a subculture collections of
characterising needs to papers merrily
what's really cool about this is that
we can start design microbial
communities based on the lean people
that we can have the that we can faded
Jim free mouse with and protected from
picking up the P.'s the of these
people's migraines and gaining weight
which would normally do when I house
with the mouse not pleased with the use
micro by so we can start to get a
transmission of my credits good
correlated with these changes in
physiological state I I central time
just gonna say you this one very a very
quick piece of why being able to
integrate the data across different
projects is really critical say
associated the scrap of or infant
development and what's us where looks
like roll with smooth progression but
what's really cool about this is when
you integrate with human microphone
project days or so know what we to look
at is exactly the same data set their
projected into space all the data from
a human microphone project with them
out the channels in the P C.s and each
each each frame this one week at the
infant because my combine at the time.
And you can say that this is true men
just change any of my crib I'm starting
a but you might expect is from some of
the data that's the the project
presenters you can see some of these
from week to week me and micro by a
much larger than any of the differences
between two healthy adults with or
human micro by project despite the fact
we're looking at the trajectory of just
one person. And one thing that's really
cool that's coming up here just a
moment as to be infant get antibiotics
for ear infection one this is
tremendous regression of the market by
oh it in this case spiro clear recovery
setup by pretty quickly so just rewind
back so you can see it again what we
see what the antibiotic administration
is list remain just regression of the
market by undoing month the bow and
followed in this case by relatively
right recovery but as a problem peter
about double a responsible same
antibiotic is totally different in
different people. And then this one
case what you can see is be ones up and
hope you don't configure later on and
yet as you can from several speakers
yesterday with marketplace remembers
and Chandler intervention in these
really like bins despite apparent chaos
can have profound affect something
decide later on in terms of things like
asthma allergies and even ABC but we
can also do the same thing for people
transplants a part of the why you might
want to place yourself almost kind of
map really clear illustrations the case
clustered in the facilitation please
say P C.'s from people would see that I
totally different from anything we see
in the HMP healthy subjects this is
what we did without streets might best
be of the university of minnesota. so
what's gonna happen is for these
patients again to get people transplant
from the same human And so then the
question is how much we gonna come to
resemble with the data community what
was marked by based there. And say I
CDF kills forty thousand people here
the US align one of the most common one
of the most common possible infections
say I'm so for them again to get people
transpired that one day to each frame
in most of the day but you can see is
essentially immediately all the
microbial communities shift and the
healthy stays and then they stay there
a sentence of diarrhoea disappear
within the day or two at night it's
ninety five percent. cases and using
this kind of track and we can just see
directly affects of these my combine
base there is So I'm saying in response
to transcribe the same kind of thing is
very promising pragmatics spree addicts
and other kinds of treatments you might
want to apply. So the ability to track
the stuff them intuitive way that you
can explain to the patient look nation
I think it's gonna be really critical
getting forward so you are so
essentially what we're trying to do is
develop restoration ecology the got
which you can think of as being kind of
like the wall we have this initially
partially "'cause" system maybe it's
ravaged by disease or antibiotics if
you just leave it here so you're
probably going to get a lot of weeds
growing there seventy this is what
should you JCDC didn't get markets of
pragmatics to deeper might be great
good Mike recipe people addicts or
should you do if we call bacteria there
and and essentially just a chance a
whole ecosystem and then the question
is which strategies what base which
people but I basically to you is would
you get many different people there and
so that's really the challenge we have
going for what how new system inside
this you might combine to specify the
statements by individual. Um and what
would happen is crap and the project I
never got to really provide seats at
many different strains we can use this
kind of project a set a thousands of
members of the public been interested
in the sub I have to find out what we
get is my ideas insidious trends that
are not everyone's equally excited
what's in their anyway what that I just
I think whatsoever. people apparently
informally in my lap it contributed
sensible stuff any if you might mention
specifically during that or there's a
lot to me collaborations and the
various sources support And finally
thanks for your attention I'd be
delighted to answer a few questions at

Share this talk: 

Conference program

Introduction of the Session 1 : The Gut Microbiome: Facts and Figures
Josef Penninger, Institute of Molecular Biotechnology, Vienna
23 Oct. 2014 · 9:07 a.m.
The role of commensal bacteria in the gut
Willem de Vos, Wageningen University, The Neterlands
23 Oct. 2014 · 9:31 a.m.
Q&A : The role of commensal bacteria in the gut
Willem de Vos, Wageningen University, The Neterlands
23 Oct. 2014 · 10:29 a.m.
Gut microbial richness impacts human health
Dusko Ehrlich, INRA, Jouy-en-Josas, France
23 Oct. 2014 · 11:07 a.m.
Q&A : Gut microbial richness impacts human health
Dusko Ehrlich, INRA, Jouy-en-Josas, France
23 Oct. 2014 · 11:44 a.m.
Cross-talk between the mucosal immune system and environmental factors
Hiroshi Kiyono, The University of Tokyo, Japan
23 Oct. 2014 · 11:56 a.m.
Q&A : Cross-talk between the mucosal immune system and environmental factors
Hiroshi Kiyono, The University of Tokyo, Japan
23 Oct. 2014 · 12:31 p.m.
Introduction of the Session 2 : Host - Microbiome Interaction
Susan Suter, University of Geneva, Switzerland
23 Oct. 2014 · 1:41 p.m.
Mechanisms of cross talk in the gut
Annick Mercenier, Nestlé Research Center, Lausanne, Switzerland
23 Oct. 2014 · 1:55 p.m.
Q&A : Mechanisms of cross talk in the gut
Annick Mercenier, Nestlé Research Center, Lausanne, Switzerland
23 Oct. 2014 · 2:34 p.m.
Relationship of diet to gut microbiota diversity, stability and health in older people
Paul O'Toole, University College Cork, Ireland
23 Oct. 2014 · 3:52 p.m.
Q&A : Relationship of diet to gut microbiota diversity, stability and health in older people
Paul O'Toole, University College Cork, Ireland
23 Oct. 2014 · 4:27 p.m.
Gut microbes and their role in malnutrition and obesity
Rob Knight, University of Colorado, Boulder, USA
24 Oct. 2014 · 9:16 a.m.
Q&A : Gut microbes and their role in malnutrition and obesity
Rob Knight, University of Colorado, Boulder, USA
24 Oct. 2014 · 10:01 a.m.
The gut metagenome - your other genome
Jun Wang, BGI, Shenzhen, China
24 Oct. 2014 · 10:19 a.m.
Q&A : The gut metagenome - your other genome
Jun Wang, BGI, Shenzhen, China
24 Oct. 2014 · 10:53 a.m.
Fecal transplant to mine for novel probiotics
Max Nieuwdorp, Amsterdam Medical Center, The Netherlands
24 Oct. 2014 · 11:04 a.m.
Q&A : Fecal transplant to mine for novel probiotics
Max Nieuwdorp, Amsterdam Medical Center, The Netherlands
24 Oct. 2014 · 11:25 a.m.
Introduction of the Session 4 : Nutritional Interventions
Keiko Abe, The University of Tokyo, Japan
24 Oct. 2014 · 12:46 p.m.
Interactions between gut microbiota, host genetics and diet
Liping Zhao, Jiao Tang University, Shanghai, China
24 Oct. 2014 · 12:56 p.m.
Pediatric intervention - what works and what doesn't work
Hania Szajewska, The Medical University of Warsaw, Poland
24 Oct. 2014 · 1:47 p.m.
Q&A : Pediatric intervention - what works and what doesn't work
Hania Szajewska, The Medical University of Warsaw, Poland
24 Oct. 2014 · 2:15 p.m.
Perspectives for nutrition and the gut microbiome
Nicholas Schork, J. Craig Venter Institute, La Jolla, USA
24 Oct. 2014 · 3:02 p.m.
Q&A : Perspectives for nutrition and the gut microbiome
Nicholas Schork, J. Craig Venter Institute, La Jolla, USA
24 Oct. 2014 · 3:46 p.m.

Recommended talks

Nutrition, metabolic health, cancer and NCDs
Elio Riboli, Imperial College London, England
28 Sept. 2017 · 2:27 p.m.