Tip:
Highlight text to annotate it
X
Richard Gibbs: Thanks, Rick, and thanks for everybody to
listening to our tag team here. This is how we see the big picture for the clinical translation
and the different applications that can be done at different scale of activities. So
we're on the left here, the smaller projects. Here is targeted clinical testing where you
might have really a predetermined reason to go in and do detailed study; clearly affected
families for example. Here's where we'd moved that out into a healthcare setting where it's
more generally available, perhaps without as many priors to beginning the study. And
here's truly ubiquitous application of our methods. The scale on the bottom here is a
log scale: thousands, tens of thousands, hundreds of thousands and eventually millions.
Now if we want to be out here you might say, "Why can't we be?" since you just heard from
Eric, and as you all know, that we're in this world right now where we're doing research,
and our $700 exome and $5,000 or so whole genome world, we're doing disease discovery
in projects of scale of around 1,000 to 5,000 samples, and indeed we're already looking
at cohorts with 10,000 samples or more. And if we follow the future, as Rick has just
pointed out, with the developing technologies, we're going to move out here and be in this
truly large scale range that we can extend the studies. But that is the research domain,
and so we're talking now all about the clinical translation, and there's something of a lag
there. So here we are more in the 1,000 to 10,000 sample range right now, ably doing
projects with a few hundred samples, and only in a few places getting this up into the thousands
of numbers.
So why is this running behind the research arena? Well it's clearly because that there
are different demands in the clinical translation space. I think you all know the CLIA environment
demands stable processes with 100 percent precision; the research, 99 percent precision,
just doesn't cut it. That adds cost and reduces the throughput. One has to deal with heterogeneous
samples, not going to the freezer to get 10,000 samples at a place, but getting patients in
one at a time. There's individual reporting and human input required at all steps. So,
there really are new challenges as part of the scaling -- the clinical genomics.
But it's not all bad news, of course. We've got good experience now with porting activities
from our larger research genomics projects into clinical labs. The optimal practice,
we've developed those in the context of these high-throughput genomics process. We calibrate
and stabilize the methods, document them, lock them down, declare we have standard operating
procedures, and then deploy those in a clinical environment. And we're seeing a few examples
of that now indeed happening. So we could say, indeed, with some confidence, large-scale
research studies do provide this foundation for clinical translation.
What are some of the actual examples of those transitions and throwing the technology over
the wall into the clinical labs? We have the whole genome and whole exome experimental
techniques at the lab, and there's also been really strong developments in the variant
calling and validation data handling methods, managing reference databases rationally and
effectively in the clinical setting to look for pathogenic variation, and bringing just
all of the other tools that we've learned for secure and proper management into process
management phases in the clinical lab as opposed to the experimental homes that they came from.
We've also been doing this, I think very effectively, in data sharing. You're hearing a lot of talk
about cloud and large network applications really that drive us through this on the research
end, and they're now bringing those into the clinical end as well.
So we have further opportunities to accelerate translation by more lessons joined from the
research area. You know, what's the current state? Well, we're seeing -- we're seeing
some of these activities going on in CLIA environments partly supported by insurance
companies but still being leveraged by these development programs. We're in a state now
a fraction of pediatric cases are being solved, about a quarter of them in our hands, but
I think, generally, that number is around there in other standards as well, and maybe
5 percent of cancer cases properly guarded by these activities. So we have a lot of -- I
guess the ceiling is high, is the point here. We have a lot of room to transfer, further
this knowledge over into our clinical environment. I'd say globally it's only about 10,000 cases
per year that are truly going through a clinical diagnostic environment, bearing all the bells
and whistles we have developed in the research environment.
So what are some more opportunities? Well, here's just a selection that come from us.
I think you can draw from Eric's talk and from Rick's talk other examples, but we certainly
have examples where sequencing a few hundred cases in a different context has shown us
that there are more genetics going on than we would have expected. We did a study on
cerebral palsy that showed that that 25 percent of the cases actually had a genetic contributor,
which begs the question: How ubiquitous are those genetic compounders in what is sometimes
regarded as not a genetic disease but may just have a few Mendelian forms, maybe there
are many more. How ubiquitous is that, and what could you learn from a much larger study
done with full clinical annotation?
Another example of that is a real one, we have a few cases now where we have, through
de novo mutation identification, recognize new syndromes. One case -- four cases with
a de novo out of 2,500 in our local clinical cohort. Now, that begs the question, "What's
the real frequency of this disease?" We're only seeing a fraction of possible cases.
There's a lot of filters. If you extrapolate out the birth rates and overall likelihood
of children that are in this class that might have the disorder, there could be 200 or so
of those cases here in the U.S. and many hundreds worldwide simply not recognized especially
as we extrapolate that. And what about building full genetic models by looking at various
syndromic conditions where we know that not every case can be solved as a simple Mendelian
case, but this is not necessarily the same as complex disease, but where we can find,
though detailed clinical annotation, deep sequencing, and adequate sizes, the kind of
genetic data that can build full oligogenic models that can deeper our understanding of
the disease pathology, doing genetics here somewhat from the ground up. So all this becomes
possible as we move into this additional clinical space.
So, where will we get samples for such things? Well, they're not exactly on the shelf like
the cohort studies and the case control studies that Eric spoke about, but they're not so
far away from us either. We certainly have private healthcare networks that are large
that are already being tapped into in these large-scale ways, and actually, Teri, you
cited some beautiful examples from other countries. I wish we could be -- I wish we could be counted
up there amongst those examples, and there are state screening programs. At the family
study level we certainly have advocacy groups and clinical treatment groups that can bring
us literally thousands of samples for these kinds of studies. The clinical clusters of
developmental delay, preterm birth, and families with multiple different cancers, et cetera,
et cetera, et cetera, also represent big opportunities to aggregate.
Another opportunity that's represented from the previous slides -- the menu on the previous
slides is to extend these Mendelian studies to fully describe the mutations in loci known
to cause rare disease. Eric showed the nice slide of the cancer genes where if you get
enough mutations, you begin to understand the function of the cancer gene. We need,
for these new rare syndromes; we need to aggregate the information of the mutational spectrum
from the hundreds of cases that exist so we can understand the relationship between the
kind of mutation and exactly how the disease manifests. We can only do that by coordinating
our denominator here.
And lastly, we can look most broadly at the sporadic cases to build our oligogenic models
in large sets of multiple patient admissions, such as regular screening for home medical
center admissions. And that would be the very outer rim of our onion skin diagram with the
largest input. And that sounds like hand waving and extremely optimistic, but it's exactly
also what was shown in Teri's slides that's going on in the U.K. and in other countries,
so we hope we can speed to there.
So that's the end of our tag team, prior to discussion, and I guess we can take questions
now, if we still have some time left in our allocation.