Transcriptions
Note: this content has been automatically generated.
00:00:00
Oh great this brilliant colours ranging
00:00:02
from blue to yellow to red. And it
00:00:05
looks very different. But what is the
00:00:07
underlying statistical technique to
00:00:09
show was that they were in fact it for
00:00:11
How do you approach just draw To show
00:00:14
that the profiles are different you
00:00:16
know I see no grace videos and someone
00:00:18
what is the underlying technique to
00:00:20
show that if the remote dimensional
00:00:23
microbe apropos individual individual
00:00:25
signature service different between two
00:00:27
people three people people are obese
00:00:30
normal whatever. sure well it makes a
00:00:32
difference and what you're trying to
00:00:33
accomplish so if in fact wanna
00:00:35
identified individual species that are
00:00:37
contributing to that difference then
00:00:39
you wanna look at differences in those
00:00:41
individual species incorrect for the
00:00:43
multiple comparisons. So many people
00:00:45
would you say individual T tests and
00:00:47
species abundance or other things. And
00:00:48
incorrectly P values based on the
00:00:50
multiple comparisons if you're just
00:00:52
interested in sort of group differences
00:00:54
kind a multivariate differences then
00:00:56
you could use multivariate techniques
00:00:57
and we've seen from the previous
00:00:59
speakers talks about the multivariate
00:01:02
techniques are used a lot. But the
00:01:04
problem with that is it's not picking
00:01:05
out the individual species that might
00:01:07
be contributing most of those
00:01:08
differences in order to get that you'd
00:01:10
have to do the you know very analyses
00:01:12
and then correct for the multiple
00:01:13
comparison yeah So you really inspired
00:01:18
so what's your idea idea how the
00:01:21
question but an open while studies
00:01:23
because in your intuition defences to
00:01:25
watch a here it is quite important. Uh
00:01:27
also we have a product you don't want
00:01:29
to be on the set well what time affect
00:01:32
how does it work as you intended to use
00:01:34
your how what was the what periods
00:01:36
between well we use the washer period
00:01:40
break up the serial correlation did the
00:01:41
time to go through this but because if
00:01:44
you if you measure the microbe I'm
00:01:45
serious serially right you're gonna
00:01:47
induced the serial correlation we could
00:01:50
influence the power you have to
00:01:52
differentiate effective one
00:01:53
intervention to another you have a wash
00:01:55
up here that ultimately you can break
00:01:58
up that correlation and empower the
00:02:00
problem with it is it'll lengthen the
00:02:02
study time also you could be putting
00:02:04
the person at risk if they're a washer
00:02:06
period and not an intervention. So
00:02:08
there are some serious issues with
00:02:09
these out of one designs in terms of
00:02:11
analysing the data make sure you're dry
00:02:13
compelling inferences so basically you
00:02:15
don't do anymore so well we it for
00:02:18
certain designs are certain trials
00:02:21
might make sense to have a washer
00:02:22
period if you treat someone with
00:02:25
essential hypertension the maybe you
00:02:26
could afford to have them opera drug
00:02:28
for some period of time. But if you
00:02:29
have some more acute condition it would
00:02:31
make sense to take someone off therapy
00:02:33
of any kind. So I think it just depends
00:02:35
on the setting the Just so we have a
00:02:40
very interested in Texas that is too
00:02:42
but we keep hearing about regulatory
00:02:45
issues seems like you're talking with
00:02:47
yeah yeah but that I well I'll tell you
00:02:50
what set of one studies what type of
00:02:53
evidence need to go for right use So
00:02:56
again it sort of depends on the setting
00:02:58
if a company is trying to make an
00:03:00
argument that they they have reason to
00:03:02
believe that their drug be re purposed
00:03:04
for condition or another going off and
00:03:06
doing another phase three trials in
00:03:08
individuals with that that's secondary
00:03:10
conditioner that second edition would
00:03:11
make sense shelling in a small number
00:03:13
of individuals which can ask actually
00:03:16
listed that might encourage the FDA and
00:03:19
other organising she's to at least
00:03:21
believe that there's utility that's
00:03:23
right and that intervention. So there I
00:03:25
think that these studies would be
00:03:27
appropriate. However you're doing some
00:03:29
you know fluoride in the water likes
00:03:31
that we want a lot that public health
00:03:33
benefits then the other one studies
00:03:35
make no sense at all. So yeah there are
00:03:36
certain settings real I would see the
00:03:38
end of one study simply are appropriate
00:03:40
in the FTAX is is aware that okay just
00:03:46
a conceptual question like the the idea
00:03:48
of phone studies peaceful solution for
00:03:50
that purpose deformities the last I was
00:03:53
okay so each that was a little before
00:03:56
is okay besides what what I it's a it's
00:04:00
a huge problem so I didn't you know
00:04:02
since this wasn't a cancer focus didn't
00:04:04
wanna go through that yeah tumour
00:04:05
header you need is a real problem in
00:04:07
treating cancer. However there are
00:04:09
people charlie's one from the UK and
00:04:11
others who tried to the determine in
00:04:14
project just where the tumour might go
00:04:17
based on re biopsies. So you know if
00:04:19
you see an additional said a mutation
00:04:21
right in and that's static to more than
00:04:26
maybe you could intervene at that
00:04:28
moment with a different route but you
00:04:29
couldn't anticipate that it I didn't
00:04:31
know that those mutations will present
00:04:33
the first place so you're always gonna
00:04:35
be one step behind one what one can do
00:04:38
is crafted framework in decide what
00:04:40
drugs to get one in fact certain
00:04:42
mutations right and then test that
00:04:44
algorithm and test its utility
00:04:46
impatience and that would be sort of
00:04:48
ended aptitude and one like design yeah
00:04:54
oh Need I think your remote I'm just
00:04:56
wondering not to be provocative whether
00:04:59
quality setting contrived is the best
00:05:01
way to we should fall right because in
00:05:04
a way to go but try to make some but
00:05:07
you would then like all right. So so
00:05:10
you see that once a clear phenotype is
00:05:13
that's the key to it is then you would
00:05:15
use the same drug treatment centre
00:05:17
dramatically without any adjustments
00:05:19
preparation is this what you're saying
00:05:22
you could so I I I do the initial exam
00:05:25
we're just doing in out of one in
00:05:26
isolation and there are settings where
00:05:29
that makes perfectly good sense in the
00:05:30
impact disease case you're not gonna
00:05:32
find any more patients probably gonna
00:05:34
have one if the the diseases to
00:05:36
cleopatra. So you want to see if the
00:05:38
drugs working you'd have to do another
00:05:39
one that's a setting where the end of
00:05:41
one as it is called for however if
00:05:43
you're gonna treat the cancer patients
00:05:46
on the basis of their original profiles
00:05:48
then the more accurate and one makes
00:05:50
sense what I wanna press one people is
00:05:52
then you're no longer testing
00:05:54
individual drivers your testing an
00:05:55
algorithm that matches the drugs that
00:05:59
to the tumour profiles. So for example
00:06:01
if there is any dietary interventions
00:06:03
associated with different features in
00:06:04
the micro by then the degree to which
00:06:07
those those different features show up
00:06:09
in anyone individuals microbe I'll you
00:06:11
might use different combinations of you
00:06:13
nutritional supplements or whatever it
00:06:16
might be but you need to have the
00:06:17
algorithm tell that invents because if
00:06:19
you wanted to show what had utility you
00:06:22
get a bunch of people you match the
00:06:24
nutritional interventions on the basis
00:06:26
of the microbe I'm profile right and
00:06:28
then maybe for some people they
00:06:30
wouldn't have any nutritional
00:06:32
intervention and see if that algorithm
00:06:34
the way you match the nutrition oh a
00:06:36
pro interventions to the profile makes
00:06:40
sense. So again which are no longer
00:06:41
testing is a specific you personal
00:06:44
intervention but away from actually
00:06:46
nutritional interventions to the
00:06:48
profiles and that's the thing that is
00:06:50
on the minds of people because it's no
00:06:51
longer the case you can factory go
00:06:53
individualised matters such that people
00:06:56
with unique gentlemen features are to
00:06:58
get a particular drug and someone
00:07:00
crafts a new drug comes up with a new
00:07:02
drug new intervention for the express
00:07:04
purpose of combating that type of
00:07:06
deleterious profile you wouldn't find
00:07:09
necessarily enough people to actually
00:07:11
test the intervention on so what you
00:07:14
have to do is embedded in sort of an
00:07:16
aggregated and one setting and see if
00:07:18
it has utility as part of a much larger
00:07:20
out. but oh how would you sort this is
00:07:24
a oh would you address safety with you
00:07:27
grow up you know part of like this. So
00:07:29
for the stand to cancer trial there's
00:07:31
at least five investigational agents
00:07:33
know these is all past kind of phase
00:07:35
one potential not to be toxic but now
00:07:38
and showing that they have efficacy
00:07:39
they're buried into the trial didn't
00:07:41
say again "'cause" it's the time but
00:07:43
that standard cancer trial we have
00:07:45
about twenty already out there marketed
00:07:48
rocks and about five or seven
00:07:51
investigational agents from different
00:07:52
from so companies who just wanted to
00:07:54
know if they're drug might have utility
00:07:56
in people with specific tumour