Tip:
Highlight text to annotate it
X
Scott Weiss: Okay. I'm not really going to talk a lot about
the mechanics of whole genome sequencing. I'm going to describe how the Partners Center
for Personalized Genetic Medicine is organized, describe the Laboratory for Molecular Medicine,
which is our CLIA-certified lab, talk a little bit about our plans to launch this service,
which is very much along the lines of what Howard and David are doing at MCW, and focus
in on Gene Insight Lab and Clinic as a critical piece of software and something that may be
sharable with other members of the group.
So how is PCPGM organized? We have a set of research cores, which are the sort of the
standard cores that are also used by our CLIA-certified lab, and these research cores both support
research and support the clinical lab, and they are CLIA-certified. We're supporting
about a $140 million of research across Partners Healthcare. Just for those of you that don't
know, Partners Healthcare is the umbrella organization for Massachusetts General Hospital
and Brigham and Women's Hospital, and involves a number of ancillary facilities that are
linked to those two academic medical centers.
So this is, this is all CLIA. We also have a very active RPDR, and electronic medical
record that's used for research. Shawn Murphy [spelled phonetically] actually runs this,
it's not actually technically part of the center. Most of the research that's being
done here is not genetic. There are about 3,000 projects that are going and over four
million Partners patients that are being used for research using the electronic medical
record. And this is the piece that's not yet built. We've just actually gotten approval
to go forward with the Partners Biorepository for Medical Discovery, which is going to be
a biorepository which will be linked to both the CLIA lab and the electronic medical record
and our goal is to have the biorepository actually CLIA-certified as well, because going
forward this, we're doing this in a consented fashion, and I think the reasons for that
are manifold, but I think the most important is is that we want to be able to do research
on these patients but also be able to push things forward in terms of clinical research
and clinical care and it's going to be a lot easier to do that if this is both CLIA-approved
and if the patients are consented.
So that's how the center is organized. So between the investigators that we're supporting
here -- currently there is a service, Crimson [spelled phonetically], which is now just
became operative at Mass General, has been operative at the Brigham for the last three
years, and that's supporting about $40 million in research. So between $140 or so million
dollars' worth of research that's using the MR, another 150 that's using these cores and
the 40 here, we're supporting about $300 million of research for approximately 400, 500 Partners
investigators across the healthcare system.
And I think -- sorry -- [unintelligible] go back. Perhaps the most important part is the
IT infrastructure that supports all this, because without this IT infrastructure, we
wouldn't be able to have samples flowing seamlessly from the CLIA lab to the research cores to
the bio-repository, to other investigators, samples going to the Broad or wherever. So
the IT infrastructure here is clearly the major investment in terms of the center and
the way it works.
So a little bit about the Laboratory for Molecular Medicine. Heidi Rehm, who was at the previous
NHGRI workshop with Robert Green last week, is the director of the lab. This lab's been
in existence since 2003 and we do about 4,000 genetic tests a year. We do these tests for
both the two Partners hospitals, but also for medical institutions around the country
and in some cases around the world. The bulk of our testing is in cancer and cardiovascular
disease. We do have some other [unintelligible] and monogenic traits that we test for and
currently we're -- we have tests for over 200 genes. The challenge here, obviously,
and this is the example of one of our first tests, both are -- two academic cancer centers
produced papers within a few months of each other looking at EGFR as a gene for -- treating
[spelled phonetically] response for small cell lung cancer, and within three months
the lab had a test up and running to test patients and we still do the bulk of the testing
for Dana Farbor [spelled phonetically], although the MGH does their own testing now for EGFR.
A little bit of context in terms of the evolution of clinical genetic testing, which I think
gives, maybe, further justification, if you will, for the approach that Howard outlined
earlier. This -- you can see along here how the evolution of testing in the lab has occurred
and, you know, we've expanded tests and expanded the number of tests in the areas that we've
been genotyping, but we're reaching the point now where it costs -- it's going to cost as
much to genotype the whole genome as it will -- in fact, the cost of the genotyping will
be less than the cost of our cardio chip [spelled phonetically] genotyping. This, the cardio
chip test, I think, costs on the order of $3,000. It's likely that a whole genome, at
least the genomic sequencing portion of that test, is going to be well below that in a
year or so.
So we're entering this logarithmic phase where whole genome sequencing is going to be relatively
cheap, but the thing that's going to lag behind has been pointing out by a number of people
already, is the analysis and the ability to build content to deliver to practitioners.
That's going to be the major challenge going forward. It's not going to be the sequencing
itself. And we anticipate that -- you know, all of our tests now are targeted next generation
sequencing tests, and we expect to be completely out of that business in two years, simply
because it's just not going to be economically viable to do it.
So our service is organized very similar to the way MCW has organized theirs. We're outsourcing
the actual sequencing to Illumina and Complete Genomics. We're going to concentrate all of
our efforts on data analysis using the existing infrastructure, the ALA -MAM [spelled phonetically],
and all of the things that Howard talked about in terms of patient work up consent, oversight,
clinical committees is all going to be in place. And we're probably going to launch
this -- we would like to launch it in July of 2012. The major factor in terms of whether
we will be able to do that or not is going to depend on our IT infrastructure, I think.
I think it's interesting to look at this from the context from the clinician's perspective.
The amount of information that's going to be generated for clinicians is truly daunting
and new information can emerge on any of these variants at any time and new forms of support
are already needed to stay up-to-date on the limited number of variants that we've identified
using the tests that -- the clinical support tools that we have now. And the infrastructure-dependent
clinical process need to be established to allow clinicians to retrieve and manage genetic
results, to link clinicians and to experts capable of determining the implications of
each of these genomic variants and keep them up-to-date.
And obviously this whole process -- we've linked to the clinical genetics programs at
the two hospitals. Mike Murray, who runs clinical genetics at the Brigham, and David Sweetser
[spelled phonetically] who's his comparable person at Mass General, are involved with
Heidi and Robert Green and me as we as we sort of formulate the plans to launch this
service. But this also, I think, creates significant opportunities for Partners Healthcare in particular
because of tools that we've already developed and would like to share with people here.
You've got this constant flow of cases that are going to the geneticists in the lab to
sign out these cases, you've got this evolving knowledge base, and the need to continuously
update the electronic medical record and report information to clinicians.
And we've developed a tool that we call Gene Insight that we use both as a report generating
engine, but is used in the laboratory as a knowledge base. So we keep information on
all of the variants that we've genotyped in the lab and we obviously use this to report
the results to the electronic medical record. And both of these, I think, are going to be,
turn out to be very important. Heidi Rehm has actually put in a U41 grant to address
the issue that people have been talking about here, which is a huge issue, which is variants
link to clinical phenotype, allele frequencies in people that don't have the phenotype, standard
ways of reporting and annotating variants, this is going to be a huge issue going forward
and if this group decides to make that an issue that people want to collaborate on,
I think that we have several people in our center who would be very interested in participating
in that.
So this is what Gene Insight looks like if you're at a terminal in your office. This
is a patient named Curious George who happens to have a cardiomyopathy, and he had a variant
that was previously classified as of unknown significance that has been now reclassified
as pathogenic. So we can update these in real time, we can push this information to the
clinicians, we can provide decision support tools through this software and help to enable
the clinicians to manage these patients and to help them seek support from genetics experts
if they need that backup and support, so. And I think that this software, although it
is clearly at the terminal end to the pipeline that Howard described, it's still software
that is potentially valuable and we'd be interested in sharing it with others and working with
others who might want to use it.
Male Speaker: Is this sitting on the RPBR [spelled phonetically]?
This application is sitting on the research that it provides?
Scott Weiss: No.
Male Speaker: No.
Scott Weiss: No, it's not sitting on the research database.
And this is just the people who have helped get this center off the ground. Yes, Debby
[spelled phonetically]?
Female Speaker: So -- should I wait or can everybody hear
me?
Female Speaker: Really loudly.
Female Speaker: All right. Scott.
Scott Weiss: Yes.
Female Speaker: Tell me what made you change from one category
to pathogenic? That's a big change from, you know --
Scott Weiss: For an individual variant?
Female Speaker: Yeah. So what was the level of evidence that
made you make that switch?
Scott Weiss: Yeah, I'm probably not the -- I'm probably
not the best, you know, I don't run the clinical lab. That's sort of like the question that
Howard deferred to David. I think that the geneticists in the lab have a whole bunch
of criteria that they use to decide whether something, a variant, changes. They look at
the literature, they're looking at a lot of different things. But it -- and -- I agree
with you that that whole issue is one that is --
Female Speaker: A big nightmare [spelled phonetically].
Scott Weiss: Well it's, well it's, I think it's not, it's
not just a nightmare, it's one where there's not clear standards and some people might
do it one way, another people might do it another too, so I think it's going to, it's
--
Female Speaker: I mean, people are beginning to look at this
because it's -- can I borrow this for a second? -- people are beginning to look at this because
it's such an important question, but I do think it is key because what has been reported
in the literature, if you go today with modern databases, what, in a family, segregated and
what was without functional studies or even some maybe not appropriate functional study
caused, or said was causal --
Scott Weiss: Well --
Female Speaker: -- will not hold up --
Scott Weiss: Yeah --
Female Speaker: -- at all [spelled phonetically].
Scott Weiss: -- well I think that, you know, there's clearly
variants that were found for monogenic disorders where the variants are in linkage to this
equilibrium with a causal variant and it was reported that that was the causal variant.
So I couldn't agree with you more that this is a huge area. I think it's an area that's
going to demand a lot of attention. What I wanted to focus on here was not so much the
process that we do that, but the fact that the software allows us to change the interpretation.
I think that's going to be important because I think that going forward, it's -- this is
not something where it's going to be set in stone.
Female Speaker: Yeah, but I think it's important also if you're
taking data from a provider, what they're providing, let's say --
Scott Weiss: Oh, yeah.
Female Speaker: -- in a CLIA setting is not necessarily the
end of the genome.
Scott Weiss: No question. And CLIA certification doesn't
mean that an interpretation is set in stone.
Male Speaker: So you're looking at a lot of the places to
reduce cost and you're doing some of the whole genome sequencing then at the CLIA labs, like
Illumina, and when you look at these at, like, 4,000 bucks and they'll do the alignment,
call the variants with something that's good enough, do you see bringing that -- I mean,
I guess I'm wondering why bring that back if you have to reinvent all the analysis questions
and all the initial lab setup. Is that a part that you really see needs to be in-house?
Scott Weiss: I think that -- you know, we have currently
two high seqs [spelled phonetically] that are being used for research sequencing. We're
not doing any of the whole -- we have started doing whole genome sequencing ourselves in
the research core. I think that whether we ultimately bring this in-house, or how much
we do at Partners and how much we rely on Illumina and Complete Genomics is a fluid
thing, it's not something that's been set in stone. I don't know that we'll -- the one
thing I'm pretty sure we're not going to do is we're not going to build the giant sequencing
facility in Waltham to sequence all Partners patients. I don't think that's, probably,
in the cards. But how much we actually do ourselves and how much we outsource, I think,
is an open question. I think for us the issue of the analysis, the interpretation, is what's
paramount.
Male Speaker: But when you say analysis interpretation,
I think that's different than what was referred to earlier, which was how you call the variants.
Your real concern with analysis then is interpretation, so they're kind of two --
Scott Weiss: No, I think, I think we would include that
as well. And I think that all of the issues that were described are issues that we are
going to consider in terms of how we do this. Nothing here is set in stone. Virtually everything
about this is a moving target. So how we decide to do this today may be very different from
how we decide to do it, you know, six months or a year from now.
Male Speaker: Scott, you may have mentioned it, but can
you clarify whether the Gene Insight is designed as a standalone tool, or is it in any way
linkable with an EMR to --
Scott Weiss: No, it's --
Male Speaker: -- talk to EMR?
Scott Weiss: It's absolutely linkable with an EMR, you
know, several other health systems actually have been using it. We collaborate with Intermountain
Health, they're using it. So it's definitely linkable with other software and other EMRs.
Male Speaker: So I had the exact same question, but that
means it's a standalone tool, because if it's not, if different EMRs can talk to it, it
means it's not tied to a specific EMR, is that correct?
Scott Weiss: That's correct.
Male Speaker: Okay.
Male Speaker: Yeah, it's really a service player. So it
reduces the number of interfaces so you don't have to build interfaces with every single
laboratory, you know, that basically this resides between the clinical laboratories
and the information systems and so you really build one interface from a clinical site to
Gene Insight and then you can interact with all the laboratories that use that. And so
it just simplifies the interface construction with otherwise, in a distributed single gene,
gene-by-gene type of thing, you have an infinite number of interfaces that you have to build.
Now, again, it may become a moot point if we all move into next-gen whole genomes, but
at least at this point it was a very pragmatic solution.
Female Speaker: Additional comments for Scott?
Male Speaker: [inaudible]
Female Speaker: Yes.
Male Speaker: It's a follow-up on your question. As all
of these modular components are proposed and developed by different people, is there any
idea that in the end [spelled phonetically] something like data.gov can [unintelligible]
technologies will be somehow [unintelligible] or [unintelligible] that all these things
used to talk to each other? Anyone has a comment on that?
Male Speaker: This is -- yeah. So this is something that
groups are talking about and specifically since this involves a lot of moving, pushing
of data around, the clinical genomics group of HL7 has specifically constituted a working
group within their working group to look at whole genome and whole exome sequencing and
how that data can be characterized. And they have other working groups that have looked
at other types of data models within this arena for family history and there are additional
working groups that are outside of the genome working group that do clinical decision support
models, but essentially all building on a common standard language so that the -- there's
relatively seamless movement of the information that you want as long as the information system
is built using that type of a communication standard, in this case an HL7 version 2.X.
Female Speaker: Okay. Thank --