Player is loading...

Embed

Copy embed code

Transcriptions

Note: this content has been automatically generated.
00:00:00
Here I really enjoyed the stimulating
00:00:02
discussions Yeah so far and hopefully I
00:00:06
can follow suit within equally
00:00:08
interesting set of discussions and what
00:00:12
I wanna do is talk to you about my
00:00:13
experience invasion in doing target in
00:00:16
clinical trials though was that
00:00:18
leverage genetics and you know makes I
00:00:20
had a lot of discussions with different
00:00:22
regulatory sees in united states most
00:00:25
notably the FTA a that's such trials
00:00:28
I'll you some examples of trials than
00:00:29
that actually made it way make their
00:00:31
way through the FTA a bureaucracy yeah
00:00:36
all of the focus on the micro by what I
00:00:38
think again we get taste for the sort
00:00:40
of study that I believe could be
00:00:42
pursued with the micro by nutrition in
00:00:45
particular here the topics I'd like to
00:00:47
talk about first all give a brief
00:00:50
overview that none of you were
00:00:52
strangers to an individualised
00:00:54
personalised Madison in nutrition focus
00:00:57
on genetic diversity and identifying
00:01:00
the genetic basis if you've infinitive
00:01:01
pick expression this'll be crucial to
00:01:04
later point something my talk summarise
00:01:08
to the degree possible much of the
00:01:10
discussion so far on the micro by was a
00:01:12
marker of health and dietary change
00:01:16
then hey might not having to do with
00:01:18
the end of one studies their design
00:01:19
implement he Shannon extensions and one
00:01:22
study then talk importantly about a new
00:01:25
directions and then once the that to do
00:01:27
with aggregated and one studies and how
00:01:30
these studies really don't test
00:01:32
interventions so much as an algorithm
00:01:34
for matching gentlemen profiles to
00:01:37
specific interventions and this is a
00:01:38
topic again that's on the minds of many
00:01:40
people in the united states of their
00:01:43
relevant regulatory agency. So here's
00:01:47
the fee of my talk the unique genetic
00:01:50
and biochemical individuality humans
00:01:53
suggested optimal nutritional demands
00:01:55
for health maintenance may have to be
00:01:57
tailored to each person is unique
00:01:58
biological exposure profile again no
00:02:01
one in this room is a stranger to the
00:02:02
sort of thing. So can question one that
00:02:05
has come up in previous speakers talked
00:02:08
is what is the role of the microbe I'm
00:02:10
an optimising health via nutritional
00:02:12
dietary manipulations how can one
00:02:15
leverage the micro by what is the proof
00:02:17
that it's actually meaningful. We could
00:02:20
ask is being like a by a lot of of
00:02:22
being the type for assessing relevant
00:02:24
changes attributable to dietary
00:02:26
manipulations could be treated as a
00:02:29
surrogate endpoint for nutritional
00:02:31
impact just requires a affordable
00:02:34
questions to be addressed to answer
00:02:36
that is what is a healthy my combine
00:02:38
could be a therapeutic target that is
00:02:41
we could manipulate species composition
00:02:43
with and they got to list some of that
00:02:46
market is therapeutic angle in and of
00:02:48
itself that as we could use legal
00:02:51
transplants to actually put in certain
00:02:53
species into the gods of others and
00:02:56
hope to listen to change. So first are
00:03:00
personalised Madison some time ago
00:03:03
nature biotechnology pass the question
00:03:06
is personalise Madison finally arriving
00:03:09
in fact textbooks written at hitting
00:03:11
the the use personalised Madison than
00:03:13
that practise for years later nature by
00:03:17
biotechnology ask the question what
00:03:18
happened to personalise minutes I think
00:03:23
what happened to personalise Madison is
00:03:24
there wasn't enough proof to show that
00:03:26
actually worked. So the question is how
00:03:29
does one has the personalised treatment
00:03:31
actually works. The same could be said
00:03:33
personalise nutrition. Now there's many
00:03:36
different vehicles advocating for the
00:03:39
user personalised interaction fact the
00:03:42
recent paper describing the growth in
00:03:45
publications alone describing neutral
00:03:47
genetics the belief you can tailor and
00:03:49
I someone's uni do you profile but
00:03:53
again the question is how does one test
00:03:55
the personalised I actually works. So
00:04:00
before I get into the design of studies
00:04:02
and then I wanna give some background
00:04:03
for why genetics is so important in the
00:04:06
construction relevant clinical trial.
00:04:09
So as many speakers appointed up before
00:04:11
the cost of you know sequencing of
00:04:13
going down tremendously making it
00:04:15
possible in this teenage to see once
00:04:17
entire human genome. And fact work with
00:04:20
them and craig better was the first
00:04:21
person to see once an individual gina.
00:04:23
And that that time it costs probably
00:04:26
five hundred million dollars now I can
00:04:28
see what's singled you know for about a
00:04:29
thousand bucks how does one identify
00:04:33
genetic variations actually influenced
00:04:35
rates and diseases that one of anyone's
00:04:37
intelligence but I wanna walk you
00:04:38
through the basic methodology here.
00:04:41
That's why can convey the results of
00:04:42
such studies often some meaningful way
00:04:45
on the later. So essentially how this
00:04:47
works as one collects the number of
00:04:49
people with the disease and without a
00:04:51
disease and sequences. And then looks
00:04:54
for variations things that separate the
00:04:56
C once between the case control. And
00:04:59
tries right then to fire variant that
00:05:02
is more frequent among the people with
00:05:03
the disease than without it certain
00:05:06
statistical criteria are met you
00:05:08
couldn't or that you've identified a
00:05:09
genetic variance that is just also you
00:05:12
did with the disease. And hopefully in
00:05:14
a car so it doesn't always work that
00:05:16
way that it's not quite as easy as that
00:05:19
might seem typical there are variations
00:05:23
that are real to real change phones the
00:05:26
biologic the significance of which is a
00:05:28
no however if what one can do. This
00:05:31
what I assume that the variations there
00:05:34
are variations on one we can use it
00:05:36
once it's actually impact the
00:05:37
functioning of certain genes then one
00:05:40
could argue that that collection of
00:05:41
where eric's seen in the cases.
00:05:43
contributes to disease susceptibility
00:05:46
So it's not the case that only comment
00:05:49
variations are likely contributed
00:05:51
disease it could be the collections of
00:05:53
where variations each with the small
00:05:56
affected but collectively of a large
00:05:58
effect could influence a disease. And
00:06:01
there's always to identify both common
00:06:03
where variance there associated with
00:06:05
disease in fact many diseases head unit
00:06:11
variance identified they're and would
00:06:12
be associated with their express. This
00:06:15
is what is known as a Manhattan plot.
00:06:17
"'cause" it reflects the Manhattan
00:06:19
skyline. But ultimately what it
00:06:22
consists of is on the Y axis the
00:06:24
strength of the association between
00:06:26
genetic variations whose positions are
00:06:28
given along the X axis. So each little
00:06:31
dot here corresponds to the location on
00:06:34
the human genome of a genetic variance
00:06:36
that is tested for association with
00:06:38
particular disease yeah the string and
00:06:41
the association is given in this
00:06:43
instance by the minus log ten P value
00:06:46
that's that a higher value indicates a
00:06:48
stronger association no statistical
00:06:51
thresholds have to be set. So that you
00:06:53
don't just identified variance that for
00:06:55
all intents and purposes are noisy and
00:06:57
not reflective of to signal. So one has
00:07:00
this past very rigorous statistical
00:07:03
criteria in order to make the claim.
00:07:04
that are very isn't fat associated And
00:07:08
that has to do with the fact that
00:07:09
you're testing thousand that tens of
00:07:11
thousands if not millions of variance
00:07:13
or association if you do a study genome
00:07:16
why today in the slightest little data
00:07:20
but it's the most up to date and to
00:07:22
offer a number of genetic variations of
00:07:24
identified who's locations are depicted
00:07:26
on each one of these little cartoon
00:07:28
representations of the genome. And
00:07:31
these are different diseases that the
00:07:32
variations are known to be associated
00:07:34
with and they all need this very
00:07:36
rigorous just make you so we can say
00:07:39
that these variations so on. local
00:07:41
association with a particular disease
00:07:43
So for example variations on from
00:07:47
someone into influence certain
00:07:50
digestive system diseases variance on
00:07:53
from us arms and say eleven and twelve
00:07:55
influence bottom intervene types so
00:07:58
some five hundred to a thousand
00:08:00
different variations then identifying
00:08:02
unequivocally associated with certain
00:08:04
diseases yeah as a result having
00:08:08
multiple variance identified that are
00:08:11
associated with any disease. It's the
00:08:14
case that some diseases are what are
00:08:15
known as apologetic that is we have
00:08:17
many many parents that can right most
00:08:21
human Tina types are influenced by many
00:08:23
genetic variance apologetic models
00:08:26
implicate many many variant speech with
00:08:28
weak effect but collectively have a
00:08:29
large effect one can characterise these
00:08:32
power eugenics affects by using
00:08:34
appropriate statistical methods. So
00:08:36
what one can ask is not are there
00:08:38
particular individual very that
00:08:39
influence a disease. But are their
00:08:41
collections of areas that influence of
00:08:43
disease suggesting again at that rate
00:08:45
might be probably agenda about
00:08:47
contribute to some of the literature
00:08:48
describing how one can do that. But
00:08:50
this does not mean that there are
00:08:52
variants with large effects just that
00:08:54
there may be a large number of variance
00:08:57
with small affects the modify the
00:08:59
influence of the gene with the large
00:09:01
artifacts most diseases are
00:09:03
heterogeneous as a result with some
00:09:05
manifestations to do a single variant
00:09:08
with a large effect. And other
00:09:10
manifestations of that disease in due
00:09:12
to a large number of Tina their lives
00:09:17
the case this is the case because back
00:09:19
local in genetic networks don't work
00:09:21
the simple linear that as you probably
00:09:24
know in fact most genetic networks and
00:09:27
replete with feedback and redundancy
00:09:29
mechanisms such that if there is one
00:09:32
particular gene that is or maybe the
00:09:34
system could make up for it but name
00:09:36
not that there's a perturbation at a
00:09:38
different site. So what is not to be
00:09:40
going on with many diseases of the
00:09:42
genetic level is there that work since
00:09:44
is a very simple that we can probably
00:09:45
not reflective of any diseases so the
00:09:47
network you can see that there's one
00:09:49
central gene that preacher is gonna
00:09:51
upset the functioning of the entire
00:09:53
network if we see this sort of
00:09:55
situation arise perspective disease
00:09:57
then it's probably monotonic simple
00:09:59
perturbation here "'cause" as the
00:10:01
collapse of the entire system leading
00:10:02
to a disease you could also be the case
00:10:05
that the networks are wired a little
00:10:07
differently such that you need to
00:10:09
mutations are variations in jeans but
00:10:12
for the system as a whole collapses. So
00:10:14
this would be consistent with what
00:10:16
people refer to as apologetic diseases
00:10:19
pathogenic diseases are such that you
00:10:22
need multiple perturbations dipper down
00:10:24
the system and lead disease. So anyone
00:10:28
disease may have margin forms a legion
00:10:30
like forms apologetic for yeah before
00:10:35
going to the the the upshot of genetic
00:10:39
heterogeneity user describe one talk a
00:10:42
little bit about the genetic diversity
00:10:43
of human populations in here I just
00:10:46
want to rehearse the out of Africa
00:10:47
about this you get I don't mean to
00:10:49
offend anyone's intelligence they just
00:10:50
like the walking through this. So get
00:10:52
some appreciation for some of the
00:10:53
statements and then make later. So
00:10:56
what's going to happen sometime ago is
00:10:58
there a number of individuals that were
00:11:00
populating Africa in these coloured
00:11:02
symbols here just represent different
00:11:03
genetic variations there were present
00:11:05
at the time I think migration out of
00:11:07
Africa some subset of people living in
00:11:10
Africa at the time decided to move to
00:11:12
the middle east. So what's important to
00:11:14
keep in mind is that it wasn't everyone
00:11:16
in Africa who ultimately made the way
00:11:18
to the middle east only some subset
00:11:21
that suggest that only some subset of
00:11:23
mutations are genetic variations were
00:11:25
present in that at the time all we need
00:11:27
a way into the middle east at some
00:11:30
point a little later some people in the
00:11:31
middle east decided to move to europe.
00:11:34
And only a subset of the genetic
00:11:36
variations that made their way out of
00:11:37
Africa into the middle east only made
00:11:40
where you're this has important car
00:11:42
once is for the contemporary standing
00:11:44
genetic variation in the population it
00:11:48
suggests that in non african
00:11:50
communities only the migrant the types
00:11:52
of you represent could also suggested
00:11:56
since there's more recent populations
00:11:57
in Europe and the middle east and of
00:11:58
course other parts of the world that
00:12:00
resulted from this migration out of
00:12:02
Africa there'd be less time for
00:12:05
selection to wash away the deleterious
00:12:08
effect the genetic variations that
00:12:10
could cause disease essentially you
00:12:12
might see certain forms disease causing
00:12:15
various a higher frequency and then I
00:12:18
in populations like maybe someday is
00:12:20
yes this is and true also because there
00:12:24
were smaller populations that founded
00:12:27
these not african communities there'd
00:12:30
be a likely greater you can see of home
00:12:32
as a gossipping. And that has to do
00:12:34
with S phenomena of random inbreeding
00:12:37
when you're in a small finite
00:12:39
population you have a small number of
00:12:41
makes to choose from so the probability
00:12:43
that you're gonna make with someone who
00:12:45
is related to some level is high if you
00:12:48
take huge population with a large
00:12:50
number of minutes than the probability
00:12:52
of that occurring smaller. So we
00:12:55
actually did a study with complete you
00:12:56
know max about five or six years ago
00:13:00
where we sampled hold you notes from
00:13:03
individuals throughout different parts
00:13:04
of the world and just ask the question
00:13:07
how diverse these you know more with
00:13:09
respect to functional content that is
00:13:11
very issues that might cause disease.
00:13:14
And this was published again a little
00:13:17
while ago first we have to make lame
00:13:20
that the variations were likely to be
00:13:22
functional that is cause disease. So
00:13:24
how could one do that if we had
00:13:25
millions of variations which in fact we
00:13:27
did sit for music computational
00:13:30
techniques and these are widely
00:13:32
available this is just according to the
00:13:34
gene and all the elements in and around
00:13:36
the gene in but are are likely to cause
00:13:39
between the dysfunction and hence be
00:13:41
functional and could lead to disease.
00:13:43
with computational techniques we can
00:13:46
make claims about genetic variation You
00:13:49
know all these parts in that you know
00:13:51
so what we did was take all those times
00:13:53
we had a disposal run these programs on
00:13:57
it and just catalogue number
00:13:59
deleterious or likely disease causing
00:14:02
mutations or variations of the present
00:14:03
almost you know and then contrast or
00:14:05
frequency across the different
00:14:07
population groups that we yeah so we
00:14:10
get this maple bar chart showing what
00:14:12
we found easier african populations we
00:14:15
have for our populations european
00:14:17
population ageing papa patient into
00:14:19
that next populations in indian
00:14:21
population and a mexican population
00:14:24
first thing you can see is that there's
00:14:26
a greater number of just genetic
00:14:28
variations in total along yeah african
00:14:31
individuals which is consistent with
00:14:33
what we knew about this migratory
00:14:36
pattern out Africa was the oldest
00:14:38
population it was the most diverse the
00:14:40
time. So we see a greater number of
00:14:43
variations. Now we see that there's
00:14:45
almost a six million variations as in
00:14:48
the average african americans african.
00:14:50
you know and that's because the way we
00:14:53
quantified the presence of the variance
00:14:56
was not comparing them to other human
00:14:58
genome that actually she T know now why
00:15:01
we do that. That's okay contrasting
00:15:03
different global populations and would
00:15:05
make sense to use as a reference a
00:15:07
genome for any one of those
00:15:09
populations. So we ask how many
00:15:11
variations were present in each one of
00:15:13
these you know that we have the
00:15:14
radiated from the change you know so
00:15:16
it's a fair comparison so we saw a six
00:15:19
million or so variance present yeah and
00:15:21
you know and substantially less in the
00:15:25
manner you know consistent with this
00:15:27
hypothesis. Now if we counted up the
00:15:29
number of their dollar here is
00:15:31
variations ever present in the genome
00:15:33
you can see that as a result of there
00:15:35
being a greater frequency of just
00:15:37
variations overall and there and you
00:15:39
know there's a greater frequency
00:15:40
deleterious are likely disease causing
00:15:42
variations in the and you know however
00:15:46
if we ask how many homicide gets Gina
00:15:50
types there in the different
00:15:51
populations that we can see that
00:15:53
despite the fact there's a greater
00:15:55
number of overall variations in the
00:15:57
african populations there's actually
00:15:58
last as a gas. And again that's
00:16:01
consistent with what we know about the
00:16:03
migratory patterns as a result of there
00:16:05
being greater homozygous genotype so
00:16:07
the non african populations. There's a
00:16:09
greater frequency and damaging likely
00:16:12
disease causing variations that are
00:16:13
homozygous no not african populations
00:16:16
what to suggest is that on a global
00:16:18
scale humans are quite diverse just
00:16:22
think that each individual might have a
00:16:23
unique mix the disease and heart
00:16:25
disease causing very very when you
00:16:28
wanna individualised treatment for
00:16:30
someone's a disease. Yeah there armada
00:16:34
genetic disease. that a correct the
00:16:36
ball through nutritional therapy Some
00:16:38
of these diseases go by the way of
00:16:41
inborn errors of metabolism many you
00:16:44
probably familiar with these so for
00:16:45
example if K use a classic if someone
00:16:48
is born with the PKU mutation the way
00:16:50
to correct potential deleterious effect
00:16:53
that station is to treat the newborn
00:16:55
with Ida there's actually a large
00:16:58
number of efforts in different
00:17:00
communities one in the Amish mennonite
00:17:02
community these diseases or screen and
00:17:04
it was also on Sir you given to the
00:17:07
idea up to the people that the have
00:17:09
been shown to have these mutations
00:17:10
insist no then retail. So again this is
00:17:13
sort of an example about obvious you
00:17:15
originally example. Now the micro by
00:17:18
and sports that'd be nice people but an
00:17:19
interesting question is you want
00:17:21
actually measure the like a mile and
00:17:23
individuals with PK you would they see
00:17:25
any differences in the Michael by
00:17:27
knowing that there are certain it's a
00:17:29
pretty pronounced nutritional
00:17:30
deficiencies. I don't you sequencing
00:17:33
experiments just to let you know a
00:17:35
little bit about what the results are
00:17:37
are just beginning to that I think are
00:17:38
the most interesting after this
00:17:40
setting. So we see points to the genome
00:17:44
to individuals severe anorexia this is
00:17:47
a colleague okay in the back row
00:17:49
responded by a channel prices also in
00:17:51
the back of the room and what we found
00:17:53
is there was a very that was more
00:17:55
frequent among anorexic if the HX two
00:17:58
that is involved in fact that
00:17:59
metabolism in may be consistent with
00:18:02
some of the food cover versions the
00:18:04
anorexic have so sequencing can shed
00:18:07
light on a whole Lou of a very
00:18:09
interesting diet related factors. So
00:18:13
one study having to do with anorexia we
00:18:17
also see once among other individuals
00:18:19
to superset variance asking that one
00:18:22
and why did these individuals lives
00:18:24
alone. So we took their genome we
00:18:25
sequenced them we try I define unique
00:18:28
variations that makes explain why they
00:18:30
had some protective mechanisms why they
00:18:32
did not a disease that they have very
00:18:34
officially beginning care capacity
00:18:37
these sorts of questions what we found
00:18:40
were in this is a hard to chart to read
00:18:42
is that there were in fact some unique
00:18:44
variations present in bowls percent
00:18:46
variance that we sequence of the
00:18:48
biological signal can see these
00:18:49
variations is still a up for debate
00:18:52
what we're working on so sequencing
00:18:54
studies Spanish have a lifetime unique
00:18:56
features that individuals have that
00:18:58
might be contributed toward to their
00:19:01
health or disease that there's one of
00:19:04
the most important studies that like
00:19:05
our bring your attention I don't know
00:19:08
how many people or where to various
00:19:10
publications like tense back calling in
00:19:12
the UK where they look to genetic
00:19:14
influences and metabolite levels like
00:19:17
metabolite levels in humans. So what
00:19:20
they found was that there were many
00:19:23
spots on the human genome we're genetic
00:19:26
variations were present in something
00:19:28
the jewels it influenced the levels of
00:19:31
metabolites found in those individual
00:19:33
these are all locations and you know
00:19:37
harbouring variants that influence
00:19:39
different metabolite levels some were
00:19:42
associated with diseases other sorts of
00:19:44
process ease. They showed quite
00:19:48
strongly that these variants were in
00:19:51
fact influencing the levels and
00:19:54
metabolites a couple were very
00:19:57
interesting so for example there were
00:19:59
twelve variance there identified it
00:20:02
influence penetrate the fan production
00:20:04
in biology that we have speaker
00:20:06
yesterday talk about the use of trip
00:20:09
yeah yeah mediating certain diseases.
00:20:14
So the upshot of this paper is
00:20:15
inherited variance influence metabolite
00:20:17
levels many these metabolites or
00:20:20
disease related in fact they could
00:20:21
explain my individuals are susceptible
00:20:23
to disease. Because these variants
00:20:25
influence the level the metabolite that
00:20:27
metabolite is then essential for
00:20:30
certain activities a leading to disease
00:20:32
if there's a depletion say that that
00:20:34
have like a sunset these metabolites
00:20:37
might also be influenced by the micro
00:20:39
by in this is crucial. But also
00:20:42
contributes to greater human
00:20:43
individuality. So jump to a different
00:20:47
topic the microbe I'm as a marker of
00:20:49
health and dietary change I'm gonna
00:20:51
summarise some of the things from the
00:20:53
literature that other speakers have
00:20:54
talked first microbiology is terrible
00:20:59
this was shown by a doctor the boss
00:21:01
suggesting that models we got twins the
00:21:04
more similar like a buyer profile Liza
00:21:06
Goddard twins suggesting that ability
00:21:09
up to the micro by a profile yeah I've
00:21:11
been involved in research looking at
00:21:12
different Michael Myers here's one
00:21:15
looking at the world microbiology these
00:21:18
this eight here is just they Herod
00:21:20
ability it goes from zero to whatever
00:21:23
percent. So anything close to gonna
00:21:24
percentage yes there is a strong
00:21:26
genetic basis for the species a
00:21:30
proposition anymore or Michael by a
00:21:33
highly significant. So this is variance
00:21:37
as we discussed earlier a good
00:21:39
identified that influence the micro
00:21:41
buyer well so here to we have to take
00:21:45
to jean and he really like group had
00:21:47
been shown to influence or we
00:21:49
associated that with my combine
00:21:50
profiles a good question then is how
00:21:54
stable is microbe I mean this was
00:21:56
discussed earlier where in certain
00:21:58
disease states the micro by is less
00:22:01
stable. And in others it appears to be
00:22:03
more stable. But a good question is but
00:22:06
since the micro wireless table or is it
00:22:09
the case that the micro by so stay
00:22:11
stable that you can change and yeah
00:22:12
obvious answer that question is no
00:22:15
because the microbe I'm is in fact
00:22:16
associated with many many health
00:22:18
related factors including the diet as
00:22:21
has been discussed that this was a
00:22:23
study published while while ago first
00:22:25
look at the Michael by mystical back
00:22:28
there. But also in humans suggesting
00:22:30
the individuals with different food
00:22:32
intake at different like a mile
00:22:34
profiles a very interesting study that
00:22:37
again wasn't mentioned by the previous
00:22:39
speakers published published recently
00:22:42
looked at the got microbiology on a
00:22:44
daily basis for two individuals. So
00:22:47
these are the micro by a profile so to
00:22:49
individuals power over an entire year
00:22:51
what was interesting is there were
00:22:53
certain points at which the microbe I'm
00:22:55
change dramatically they could be
00:22:57
attributed to a particular "'cause"
00:22:58
this individual lived abroad and you
00:23:00
can see that the microphone and change
00:23:02
substantially the other individual
00:23:04
apollo had a diary real illness and
00:23:07
that course substantially change that
00:23:08
individuals Michael by one thing that
00:23:11
was interesting from the study is
00:23:12
because they had three ring sixty five
00:23:14
data points but the microbiology. And
00:23:17
they also kept track of dietary or food
00:23:19
intake. They could correlate certain
00:23:21
nutrients with my combine profiles and
00:23:25
found a number of significant
00:23:26
correlations suggesting that food
00:23:28
intake actually did influence them I
00:23:31
combine in single subjects another
00:23:35
interesting study had to do with the
00:23:37
diversity of the species that populate
00:23:39
again. But the stability of them motel
00:23:42
in the same individuals this suggests
00:23:46
that although there might be species to
00:23:48
develop some product to the formation
00:23:50
of the metabolite there could be other
00:23:51
species also contribute to the
00:23:53
formation of the metabolite such that
00:23:55
you could have a say metabolite level
00:23:57
despite the fact you have differences
00:24:00
in the species that a really I we also
00:24:03
for earlier about the influence of
00:24:07
endogenous processes that influence
00:24:10
metabolite levels. So it could be that
00:24:13
humans ability to synthesise an certain
00:24:15
cab light also influences metabolite
00:24:18
levels in suspicion in a few a papers
00:24:23
over the years use one by Gary squeeze
00:24:25
that former colleague of mine fields
00:24:26
strive to show that at least for if and
00:24:30
levels there were two sources there was
00:24:32
the source of the to that and being
00:24:36
synthesised in the Lever and also I'm
00:24:40
sure the fancy produced by the species
00:24:42
that populate the guy in rats and they
00:24:44
found a number of metabolites ones that
00:24:46
this is the case suggesting that again
00:24:49
there was an endogenous and exact
00:24:51
source for at least on account. like
00:24:53
creating variation that might be quite
00:24:56
broad minded individual basis and again
00:24:58
this is consistent with what we know
00:25:00
about genetic influences a metabolite
00:25:02
levels as I described earlier. So what
00:25:05
may influence metabolite levels are of
00:25:08
inherited genetic factors plus
00:25:10
acquiring by law So you can imagine the
00:25:13
following situation someone is a
00:25:15
genetic variation because of them to
00:25:17
have Lauren average for the levels they
00:25:20
don't have the appropriate I would
00:25:23
include trip if and they can't make up
00:25:25
for that this deficiency and hands that
00:25:27
could lead to disease. They could be
00:25:29
corrected by giving the person more
00:25:31
trip trips the latest installation of
00:25:35
the human microbiology project is gonna
00:25:37
collect longitudinal data to try to put
00:25:38
this together as many you know and this
00:25:41
is being pursued for a couple of
00:25:42
different diseases. So the longitudinal
00:25:44
data really tell us a lot about how
00:25:47
changes the micro by might be impacted
00:25:50
right dietary changes and vice versa so
00:25:54
that mind I wanna talk about how one
00:25:55
can design studies just sorta relate
00:25:57
dietary changes changes in the micro by
00:26:01
now focus on end of one studies in
00:26:03
describing the through a simple
00:26:05
example. So let's say that you wanna
00:26:08
treat someone for hypertension you
00:26:11
measure their what pressure you down
00:26:13
some sort of baseline you then provide
00:26:16
a banana pretences even yeah the drug
00:26:20
appeared to work a normal blood
00:26:22
pressure you might watch the law that
00:26:24
drug and give them another drug. And
00:26:26
see a lower blood pressure in this some
00:26:28
number of times. So the end of the day
00:26:30
you could say objectively whether or
00:26:32
not that drug actually more you could
00:26:35
imagine another individual who was
00:26:39
created in exactly the same way we had
00:26:40
a different profile such that a
00:26:43
different computer work leading it. So
00:26:46
the bottom line is in designing these
00:26:48
studies the idea is to try to find the
00:26:50
optimal treatment for that individual
00:26:52
in an objective as way possible yeah
00:26:56
when design these studies using all the
00:26:58
stitches technology there's been thrown
00:27:00
at large scale phase respect. So you
00:27:04
can leverage randomisation you could
00:27:07
use washout up here you could blah well
00:27:10
the patients and the subjects to what
00:27:12
drug around to be as objective as
00:27:13
possible you could use sequential or
00:27:15
their designs and you could also try to
00:27:18
assess multivariate being types that
00:27:20
was a number of issues associated with
00:27:22
these studies the most pronounced is
00:27:24
serial correlation between the
00:27:26
observations since you collectively
00:27:28
observations over time we have account
00:27:29
for the fact could imagine time one is
00:27:32
correlated with the measure time to
00:27:33
don't necessarily have to do that in a
00:27:35
lot population based whilst there could
00:27:38
be carryover that's sofa takes some
00:27:40
time for drug would be washed out of
00:27:42
the body you can switch to a person
00:27:45
over two different drummer facts of the
00:27:46
previous drug could be lingering and
00:27:49
"'cause" the phenotype to look is it to
00:27:51
make it look as though the second right
00:27:53
was in fact having some effect there's
00:27:56
a few other issues that there's a
00:27:58
number of extensions one I'm gonna
00:27:59
highlight later having to do with
00:28:00
aggregated and of one trials kind of
00:28:03
one trials have been around for a
00:28:04
while. And have in fact pursue for a
00:28:07
number of different diseases there are
00:28:10
issues in their design all this talk
00:28:11
about these now you could completely
00:28:14
randomised assignment such that the
00:28:16
order in which the drugs are given or
00:28:19
the interventions are given is right
00:28:21
reminds you to block random assignment
00:28:24
random assignment of the of a different
00:28:26
sort non random assignment you could
00:28:29
also ask question how many periods
00:28:31
would you need in which you actually
00:28:33
administer the different interventions.
00:28:35
So should you use five periods followed
00:28:37
by factories than other drug we're go
00:28:39
to to to to these can all be discussed
00:28:44
one thing that we've done is look to
00:28:46
see how efficient different designs are
00:28:48
with respect and one studies asking the
00:28:51
question what is the power these
00:28:53
studies if one decided to use different
00:28:55
design. So maybe we could try a only
00:28:58
two alterations of a particular drive
00:29:01
or maybe four and we could ask the
00:29:03
question again how powerful these
00:29:05
designs would be in what we could show
00:29:07
is that if you break out the studies in
00:29:10
certain ways. I could have greater
00:29:13
power. There's different applications
00:29:15
for of one studies again some of these
00:29:17
I've been brought to the attention of
00:29:18
the FDA centre up the being discussed
00:29:22
at a recent conference a couple years
00:29:24
ago one use that is being discussed
00:29:29
atlanta. I has to do with using them to
00:29:32
re report constructs many drugs is you
00:29:35
know Diane phase two because there
00:29:37
don't show the the appropriate human
00:29:39
biological relevance. So if you had
00:29:41
individuals we you call to a large
00:29:44
amount of we need to pick data you
00:29:45
could come to the conclusion that the
00:29:47
drugs actually having a by logical
00:29:48
effect. So these designs are actually
00:29:51
gaining traction with the number of
00:29:53
companies as well as we have here yeah
00:29:57
yeah to collect the appropriate data to
00:29:58
make the claim that your interventions
00:30:01
actually in some sort of the fact that
00:30:03
could be evaluating the micro by I'll
00:30:05
but can also be evaluating all sorts of
00:30:06
other phenotype for which modern
00:30:08
wireless devices could use. So for
00:30:10
example you could use a continues
00:30:13
glucose monitor to monitor levels
00:30:15
insulin in diabetics change different
00:30:18
treatments and see if in fact
00:30:19
treatments have a better or worse in
00:30:21
fact I can direct six you could use
00:30:24
weight gain scales different ways of
00:30:28
assessing a body mass not to see if in
00:30:31
fact interventions having any effect
00:30:33
whatsoever a BC activity levels are
00:30:37
obvious I think pursue hypertension
00:30:40
discontinuous blood pressure monitoring
00:30:42
could be used I believe that the real
00:30:45
value in these studies just collect as
00:30:47
much information as possible including
00:30:49
the mood and anxiety another levels
00:30:51
psychological profiles of individuals
00:30:53
undergoing these trials there's a lot
00:30:56
of motivation to pursue these trials
00:30:58
many people are actually tracking their
00:31:01
health status using such devices the
00:31:03
whole movement quantified self movement
00:31:06
other leverages these devices to make
00:31:08
claims about their health you actually
00:31:11
conducted through these trials of this
00:31:13
could you one having to do with the
00:31:15
benefits of use we randomised two
00:31:19
different treatment wanted directly and
00:31:21
one an ace inhibitor it looks as though
00:31:24
the ace inhibitor lower blood pressure
00:31:26
to a greater degree that than the
00:31:28
direct in this example however that
00:31:31
wasn't quite the case because on the
00:31:33
second administration oh the operation
00:31:37
here are the individual lost a large
00:31:39
amount of weight. So we couldn't
00:31:40
attribute the reduction in blood
00:31:42
pressure to the actual trial rather to
00:31:44
weight loss. We actually use these
00:31:47
studies in gene based trial not so long
00:31:50
ago I worked as a couple of former
00:31:52
postdoc said mine to sequence the
00:31:54
genome of the girl within thinking
00:31:56
about the condition get a condition
00:31:57
that the fighters you guys analysis. So
00:31:59
we sequence the genome to see if we
00:32:01
couldn't find a mutation could explain
00:32:03
or disease condition we did that and
00:32:06
found a couple of variance that looked
00:32:09
distill it might explain very unique we
00:32:11
need to pick features they have this
00:32:13
was a girl who'd been confined to a
00:32:14
wheelchair for about fifteen years and
00:32:16
had very sincere a neuromuscular a
00:32:18
defects. We ultimately tested the
00:32:21
functional consequences of this
00:32:23
variation using a bunch of different
00:32:25
construct and convinced ourselves that
00:32:27
this might be pathogenic variance
00:32:29
running or disease. It turned out that
00:32:31
there was a drug that actually targeted
00:32:33
the protein was harbouring visiting. So
00:32:35
what we did was actually measure
00:32:37
symptoms over a period of a year and
00:32:39
administer the draw and what we found
00:32:41
was that when we administer the drug
00:32:43
symptoms seems just aside. So true and
00:32:46
of one study we did use different drugs
00:32:49
in fact we found that in certain
00:32:51
settings wanna drugs given it actually
00:32:53
exacerbated the symptoms. So on the
00:32:56
basis of identifying a gene that
00:32:58
appeared happening for this little girl
00:33:00
symptoms we're gonna come up with the
00:33:01
drug and then tested out super actually
00:33:04
alleviated her symptoms now the
00:33:07
different variations on these sorts of
00:33:09
studies one can use a sequential
00:33:12
designs to see if in fact you can
00:33:14
identify the change in a shorter
00:33:16
possible but I'm not so you could
00:33:19
ultimately identified a time at which
00:33:22
the the intervention appeared to be
00:33:23
having an effect then what you could do
00:33:25
to convince yourself that effect was
00:33:27
attributable to the intervention you
00:33:28
take them a the intervention. And see
00:33:32
if the removal of the intervention
00:33:34
actually cost progression. And thereby
00:33:36
establish causality such that if you
00:33:38
got the money intervention probably
00:33:40
would still I have recreation of the
00:33:42
symptoms that you not the the symptoms
00:33:45
may never turn yeah there's also an of
00:33:48
one monitoring studies where one needs
00:33:50
to establish population versus personal
00:33:53
threshold I give an example of this is
00:33:55
actually from a project I'm involved in
00:33:57
that you can read about if you want a
00:33:59
tear project at work but there is this
00:34:01
if you measure the levels of a
00:34:04
particular by more this discreetly
00:34:05
species in the got like providing let's
00:34:07
say and you have an individual that's
00:34:09
right well and what's a high levels of
00:34:12
the species content or consistent to
00:34:13
disease. Then what might happen is you
00:34:16
could establish a population structure.
00:34:18
So that anyone who has levels that by a
00:34:21
marker greater than this partial are
00:34:23
likely to have disease to the point is
00:34:25
to keep someone away from that
00:34:26
threshold but you can't measure. This
00:34:30
file marker over time you see
00:34:32
fluctuations in the level that by more
00:34:34
so it's that if you press in this a
00:34:38
long enough time period you could
00:34:40
establish a personal threshold with
00:34:44
errors. And whatnot. And it in fact
00:34:48
this person started deviate going
00:34:50
forward from this kind of personal
00:34:51
average then you could argue that maybe
00:34:54
they're moving towards a disease state
00:34:56
you know we have the by mark a level
00:34:58
does not hear the population threshold.
00:35:00
So with longitudinal data you might in
00:35:03
fact be able to establish a personal
00:35:05
threshold. And thereby look really
00:35:09
signs of disease. Okay discuss this in
00:35:13
the literature in the context to
00:35:15
genetic susceptibility they won't go
00:35:17
through this but the second time anyone
00:35:19
jump into these aggregated end of one
00:35:21
studies I think this is where all the
00:35:22
action is what are activated and one
00:35:25
studies very simple an orientation you
00:35:27
just conduct anyone studies and a
00:35:29
number of different people and
00:35:30
establish who might be responding or
00:35:32
not responding with particular
00:35:33
intervention you just before that so if
00:35:36
you do enough of on the people
00:35:38
categorise responders another spotters.
00:35:40
And then look for patterns in those
00:35:43
individuals Gina types or Michael
00:35:45
combine the I correspond who is in is
00:35:48
in responder. So there's no reason why
00:35:50
you can't pursue sort of Meta analyses
00:35:52
on the results of these out of one
00:35:53
study yeah the biggest challenge in in
00:35:59
pursuing and of one studies is to in
00:36:01
fact makes sense to them in Edward and
00:36:03
there is a way to pursue this and has
00:36:05
to do with the creation algorithms for
00:36:07
matching interventions to patient
00:36:09
profiles. So typical clinical trials
00:36:12
test a specific intervention right.
00:36:15
They select individuals on the basis of
00:36:18
a wide variety factors maybe my combine
00:36:20
profile but maybe clinical profile and
00:36:23
the goal is to make claims about the
00:36:24
utility that intervention. So
00:36:26
alternative if there is a wide variety
00:36:28
of potential interventions like dietary
00:36:30
manipulations is to think of using
00:36:34
those different interventions one
00:36:35
appropriate in actual clinical trial
00:36:38
setting. So the probably see if there
00:36:40
are people who get an intervention that
00:36:41
matches their profile and ultimately do
00:36:44
better. So inappropriate trial in this
00:36:47
context would not that a particular
00:36:49
intervention. But an algorithm for
00:36:52
matching interventions to the patient
00:36:54
profile is actually done this in the
00:36:56
context of cancer or quickly go through
00:37:00
this. So is many you know their
00:37:02
particular drivers if administered the
00:37:04
patients with particular tumour
00:37:06
perturbations drawn seem to do better.
00:37:08
So if you have the Philadelphia promise
00:37:11
presently have leukaemia think leaned
00:37:13
back is a drug you should be given if
00:37:15
in fact you have EG power over
00:37:17
expression in your tumour in should be
00:37:18
given in media far inhibitor. So the
00:37:21
ways of matching drugs but a particular
00:37:23
tumour perturbations. However this is a
00:37:26
problem because most people's tumours
00:37:29
have multiple perturbations in the the
00:37:31
might be causing the tumours to grow.
00:37:33
So what you need to do is match
00:37:34
multiple drugs for certain combination
00:37:36
therapies to the tumour profiles in
00:37:38
order to make them more and the
00:37:41
combination of variations that anyone's
00:37:43
to remind have might be totally new
00:37:44
wants to make it a different
00:37:46
combination of drugs anyone cancer
00:37:48
page. This is actually a worn out of
00:37:52
the following example of a rather
00:37:54
infamous example this person had messed
00:37:57
like melanoma they could be attributed
00:37:59
to a particular genetic perturbation or
00:38:02
so it was thought there is a drug out
00:38:04
there that targeted that particular
00:38:05
perturbation and when administered to
00:38:07
the patient melted away the tumours.
00:38:09
However eight weeks later the choice in
00:38:12
back in the patient died because there
00:38:14
was clearly other perturbations present
00:38:16
in the tumour that accounted for when
00:38:18
they administer the drug. So this to
00:38:20
overcome one would need matching
00:38:23
multiple perturbations in the patient's
00:38:25
tumour with a wide variety of drugs.
00:38:28
And the different rules we're doing
00:38:30
this I'm not gonna go through them
00:38:31
simple rules match runs based on
00:38:33
pathways you can try to predict drug
00:38:36
responsiveness using indeed your mouse
00:38:38
models you can use something called
00:38:40
time to think that to up with room
00:38:42
where you can use integrated approaches
00:38:45
the basic strategy for this is you have
00:38:47
a patient and you don't know how to
00:38:48
treat them. So you draw or extract
00:38:51
material and you come up with some
00:38:53
profile the microphone patient's
00:38:55
pathology this could be a micro by it
00:38:57
could be sequencing the genome it could
00:38:59
be gene expression profile on the
00:39:01
tumour you have that signature and you
00:39:04
go to say in any case where the people
00:39:06
have been administered certain drugs or
00:39:09
interventions or combinations of
00:39:10
interventions in email the outcome. So
00:39:13
then you take your patience profile and
00:39:14
see it matches any in the database if
00:39:17
it does then you read off what
00:39:18
intervention it might make sense to
00:39:20
provide for that particular patient and
00:39:22
hopefully administer that intervention
00:39:25
to the patient interact through the
00:39:27
disease. We're actually given a pretty
00:39:30
sizable talk to change to look at this
00:39:33
in the context of cancer having to do
00:39:35
with a unique form of melanoma Colby
00:39:38
ref while I don't know so these are
00:39:40
just some of the team members but the
00:39:42
idea was to in fact test a method for
00:39:46
matching drugs to the chamber profiles.
00:39:48
So the important thing once we were
00:39:50
actually batting an algorithm for
00:39:52
matching the interventions that to the
00:39:55
unique a patient profiles one question
00:39:58
that came up that's being debated still
00:40:01
by the FBI officials has to do with the
00:40:04
ethics of actually conducting a study
00:40:07
where you've batting an algorithm like
00:40:08
this. What we could have done in this
00:40:11
context was just register patient and
00:40:14
then randomise them to either getting
00:40:16
the drugs that master tumour profile or
00:40:19
get the standard of care. So after the
00:40:21
randomisation they either get drug
00:40:24
match them or standard here. This was
00:40:26
thought to be an inefficient design
00:40:28
because if the creation had a profile
00:40:31
there wasn't obvious what combination
00:40:33
drug give the but they're randomised to
00:40:35
getting the drug that master to profile
00:40:38
than that would water down the signal
00:40:40
associated with the you know he died
00:40:44
there and lead to the possibility that
00:40:47
wouldn't see a difference between the
00:40:49
group they got that you a profile and
00:40:51
that you know what we guided drives
00:40:52
versus the standard treatment. So what
00:40:54
the FDA insisted on with the design
00:40:56
like file rewrites location we do the
00:40:59
genome profile yeah we can actually
00:41:02
identify a match. So a set of drugs
00:41:05
that here the to to be able to combat
00:41:09
the unique profile in the patience to
00:41:11
more than they were rolled in the
00:41:13
study. I really samplers that they had
00:41:17
a profile that might benefit from
00:41:18
certain combinations of drugs then they
00:41:20
were randomised together getting
00:41:23
originally guided therapy or standard
00:41:26
of care. And the reason that this was
00:41:28
complicated quickly as if in fact the
00:41:31
position knew that there was a match
00:41:32
there then they wouldn't wanna see
00:41:34
their patient randomise the standard of
00:41:36
care. So there's a lot of they still
00:41:39
growing debate about whether this is
00:41:41
the most appropriate trial design to
00:41:43
actually that algorithms for matching
00:41:45
interventions to patient profiles
00:41:47
rather than testing unique
00:41:49
intervention. So they're important
00:41:52
questions in these sorts of aggregated
00:41:54
designs what constitutes a tumour
00:41:57
profile intervention match is it just
00:41:59
just score whatever low ranking matches
00:42:01
chosen be down wait the patient in your
00:42:04
analysis what if the available
00:42:06
information interventions might be lit
00:42:08
a limited making the matches difficult
00:42:11
what about low frequency profiles
00:42:12
people that have very rare Michael by
00:42:15
profiles or tumour profiles what you do
00:42:17
with that in your try what asked a
00:42:19
should be used to establish the profile
00:42:23
you might need to consider combinations
00:42:25
of interventions that it never been
00:42:26
that show to have utility before it
00:42:30
would be crucial never crossover
00:42:31
mechanism if the therapy ultimately
00:42:34
settled on did not appear to work out
00:42:37
of the box. But of course the knowledge
00:42:39
base will increase over the course of
00:42:41
the trial. So you could ultimately
00:42:42
design trials to that these algorithms
00:42:45
for matching interventions to to
00:42:48
patient profiles in that way. However
00:42:52
there is one real motivation for these
00:42:53
done and that's one that's been
00:42:56
discussed a lot recently. I again most
00:42:59
trials has been intervention in the
00:43:00
population at large using C standard
00:43:03
faced retrial designs focus an average
00:43:06
of text and the population we heard
00:43:07
number of trials in the previous
00:43:10
speakers talks about so try their all
00:43:15
the comparator intervention or placebo
00:43:16
being used in a very regimented open on
00:43:19
real world setting the main idea is to
00:43:22
get the intervention determine its
00:43:23
utility going for or with future
00:43:25
patients right. So the subjects in the
00:43:29
trial are essentially sacrificing
00:43:30
themselves for the benefit of the
00:43:32
future patients so they're sort of
00:43:34
being treated as in it that's is true
00:43:35
of most things retrial especially when
00:43:37
it to see that was used however with
00:43:40
these and one trial the goal is to
00:43:42
focus on the individual's health. And
00:43:45
determine what might be optimal for
00:43:46
that particular patient. Now in the
00:43:48
process you gonna learn a lot of
00:43:51
interesting biology but what would be
00:43:52
correlated to work that patient at the
00:43:55
end of the day you'll come up with
00:43:56
something the page you can benefit
00:43:58
right then in there. And that is unique
00:44:00
in these sorts of spent okay so was not

Share this talk: 


Conference Program

Introduction of the Session 1 : The Gut Microbiome: Facts and Figures
Josef Penninger, Institute of Molecular Biotechnology, Vienna
Oct. 23, 2014 · 9:07 a.m.
648 views
The role of commensal bacteria in the gut
Willem de Vos, Wageningen University, The Neterlands
Oct. 23, 2014 · 9:31 a.m.
568 views
Q&A : The role of commensal bacteria in the gut
Willem de Vos, Wageningen University, The Neterlands
Oct. 23, 2014 · 10:29 a.m.
145 views
Gut microbial richness impacts human health
Dusko Ehrlich, INRA, Jouy-en-Josas, France
Oct. 23, 2014 · 11:07 a.m.
354 views
Q&A : Gut microbial richness impacts human health
Dusko Ehrlich, INRA, Jouy-en-Josas, France
Oct. 23, 2014 · 11:44 a.m.
Cross-talk between the mucosal immune system and environmental factors
Hiroshi Kiyono, The University of Tokyo, Japan
Oct. 23, 2014 · 11:56 a.m.
331 views
Q&A : Cross-talk between the mucosal immune system and environmental factors
Hiroshi Kiyono, The University of Tokyo, Japan
Oct. 23, 2014 · 12:31 p.m.
Introduction of the Session 2 : Host - Microbiome Interaction
Susan Suter, University of Geneva, Switzerland
Oct. 23, 2014 · 1:41 p.m.
143 views
Mechanisms of cross talk in the gut
Annick Mercenier, Nestlé Research Center, Lausanne, Switzerland
Oct. 23, 2014 · 1:55 p.m.
393 views
Q&A : Mechanisms of cross talk in the gut
Annick Mercenier, Nestlé Research Center, Lausanne, Switzerland
Oct. 23, 2014 · 2:34 p.m.
106 views
Relationship of diet to gut microbiota diversity, stability and health in older people
Paul O'Toole, University College Cork, Ireland
Oct. 23, 2014 · 3:52 p.m.
241 views
Q&A : Relationship of diet to gut microbiota diversity, stability and health in older people
Paul O'Toole, University College Cork, Ireland
Oct. 23, 2014 · 4:27 p.m.
Gut microbes and their role in malnutrition and obesity
Rob Knight, University of Colorado, Boulder, USA
Oct. 24, 2014 · 9:16 a.m.
1377 views
Q&A : Gut microbes and their role in malnutrition and obesity
Rob Knight, University of Colorado, Boulder, USA
Oct. 24, 2014 · 10:01 a.m.
The gut metagenome - your other genome
Jun Wang, BGI, Shenzhen, China
Oct. 24, 2014 · 10:19 a.m.
157 views
Q&A : The gut metagenome - your other genome
Jun Wang, BGI, Shenzhen, China
Oct. 24, 2014 · 10:53 a.m.
103 views
Fecal transplant to mine for novel probiotics
Max Nieuwdorp, Amsterdam Medical Center, The Netherlands
Oct. 24, 2014 · 11:04 a.m.
735 views
Q&A : Fecal transplant to mine for novel probiotics
Max Nieuwdorp, Amsterdam Medical Center, The Netherlands
Oct. 24, 2014 · 11:25 a.m.
Introduction of the Session 4 : Nutritional Interventions
Keiko Abe, The University of Tokyo, Japan
Oct. 24, 2014 · 12:46 p.m.
108 views
Interactions between gut microbiota, host genetics and diet
Liping Zhao, Jiao Tang University, Shanghai, China
Oct. 24, 2014 · 12:56 p.m.
465 views
Pediatric intervention - what works and what doesn't work
Hania Szajewska, The Medical University of Warsaw, Poland
Oct. 24, 2014 · 1:47 p.m.
265 views
Q&A : Pediatric intervention - what works and what doesn't work
Hania Szajewska, The Medical University of Warsaw, Poland
Oct. 24, 2014 · 2:15 p.m.
Perspectives for nutrition and the gut microbiome
Nicholas Schork, J. Craig Venter Institute, La Jolla, USA
Oct. 24, 2014 · 3:02 p.m.
1297 views
Q&A : Perspectives for nutrition and the gut microbiome
Nicholas Schork, J. Craig Venter Institute, La Jolla, USA
Oct. 24, 2014 · 3:46 p.m.

Recommended talks

Alzheimer's disease: genome-wide clues for novel therapies
Rudolph E. Tanzi, Massachusetts General Hospital, Charlestown, USA
Oct. 22, 2015 · 3:15 p.m.