Female Speaker:
The topic today is next generation sequencing in the clinic, gene panel testing for inherited
conditions. And I know you’ve all been, you know, listening
to many NGS talks already, so I don’t have any technical background, but I do have a
couple of background slides just to state facts as a prelude to what I’m going to talk
to you about. So, this is a very, very widely publicized
slide showing the decline of costs per — here for a genome caused by the advent of next
generation sequencing. And just to refresh everybody’s memory, it’s
been fairly recent that NGS, you know, started being available first, of course, in research,
but that was about 2005. And very, very shortly after, it actually
started being implemented in the clinics. So it was very — a couple labs in 2009. My lab here implemented in 2011 and that was
still considered early. So, you can see that the cost declines more
and more, and as a result, it is now increasingly implemented in many, many laboratories. The majority, to the best of my knowledge,
focus on gene panels. However, implementation of exome genome sequencing
is also quickly increasing, and I’ll touch on that a little bit towards the end. And here is why NGS has been so successful
and so interesting and attractive for many. So, what you see here is an example of how
our detection rates for our diagnostic gene panels evolved over time. So, you see the percent of positive cases
on the Y axis. And then, going across, different gene panels
that our laboratory here was offering at different time points starting very early in 2006 or
2007 with just five genes, and then moving up to 10, 19, 24. And NGS started happening over here with,
at the beginning, 46 genes. And the point I’m trying to make here is that
over this timeframe, we nearly quadrupled our detection rate. And this was for dilated cardiomyopathies. The scale is a little bit unfortunate because
it doesn’t look as much, but you go here from about 10 percent to nearly 40 percent. So, that’s why everybody got so excited. It seems like a great thing to increase detection
rates. And now the obvious question is is that a
good thing for every disease? When should one do this type of testing? Is it a generally applicable? So, I thought the best thing would be to use
a real disease example. And of course, I chose the one that I’m most
familiar with — [laughter] — which is inherited cardiomyopathy. So, what you see here is just a, you know,
a representation similar to common cardiomyopathies. We have hypertrophic dilated arrhythmogenic
right ventricular and a couple of other rare ones. And collectively, their incident is about
— is greater than one in 500 individuals. They all are quite severe. They can lead to sudden cardiac death, and
what they all have in common, in addition, is that they have a substantial genetic component. And all this makes up for a really high incentive
for predictive testing — diagnostic and predictive testing, I should say. But why screen for mutations? And that should be obvious, but I think it’s
useful to review. So, for diseases like — for all genetic testing,
there is, in principle, two different incentives: One, clinical management; and that is sort
of the lesser of the two reasons for cardio myopathy, but there are some examples. For example, there’s a cardiac variant of
a rare disease that manifests only as hypertrophic cardiomyopathy. And now, if you don’t test for this, you know,
using your gene panel, you can actually pin point the cause to the fabry gene, and then,
potentially give enzyme replacement therapy and cure this individual. But by and large, clinical management is not
quite yet — quite as available for cardiomyopathy as it is for other disorders. Cost is another argument, though. So, current guidelines recommend clinical
screening of first degree relatives, of affected first degree relatives. And here is something from a, you know, five
year old paper from Carolyn Ho, but it’s still relevant. But for the child of a patient with hypertrophic
cardiomyopathy, this recommended clinical screening, it comes to $6,000 for puberty
and $20,000 over their lifetime. And if you contrast that with genetic testing,
you can see how that starts saving money because once you have a pathogenic variant, you no
longer have to follow up every first degree relative. You can just reduce that to mutation positive
family members. And just to state the same thing with a more
recent paper, this is from our laboratory as of this year reporting our experience with
nearly 3,000 probands with hypertrophic cardiomyopathy. What we did here is to look at the impact
of identifying positives and no longer having to screen negatives, affected negative individuals. And so, that came out to be about $1.7 million
savings over that cohort. So anyway, that’s just meant to sort of set
the stage as to why genetic testing is believed to be useful for these disorders and showing
a little bit the history of how it has evolved. It’s actually overall a very young discipline
for cardio myopathy. It started in 1990 only, where the first gene
was discovered. And you know, 13 years later, we had the first
test, the very small test, as I showed you before. It was just a couple of genes and then moving
up quickly to our next generation sequencing in 2011. And now, it is routine to screen even more
than 51 genes for patients. It’s rapidly expanding, as you probably know. So, what makes — so, there are challenges. And what are those? That gets us a little closer as to what disorders
really benefit from next generation sequencing. The cardiomyopathies all have locus heterogeneity,
and what we mean by that there is one disease, but the mutation can reside in any one of
several to many genes. We also have allelic heterogeneity, meaning
that there are usually many different disease causing variants in a given gene, to the extent
that the majority can be private. We also don’t quite understand the pathogenic
variation spectrum yet because it’s a young discipline and because there’s so many new
mutations that arise. So, typically, you know, what — to sequence
it is — or for a very long time many, many thousand probands, you will eventually understand
the spectrum of prevalent pathogenic variants. We’re not quite there yet for cardiomyopathies. And then finally, there is a fair deal of
phenotypic clinical overlap, which can complicate the testing process. What I’ll do in the subsequent slides is move
through examples for each one of these [unintelligible]. So, there’s a bunch of them — Female Speaker:
A photo or anything — Female Speaker:
Is it possible to mute — I get a lot of background noise. Would it be possible to mute? Thank you. Okay, so locus heterogeneity was the first
problem and that’s illustrated here for dilated cardiomyopathy on the left and hypertrophic
on the right. And as you can appreciate, it’s not terrible. For HDM, it is actually quite manageable. But still, I mean, you have two main genes
and then a whole slew of other genes that can contribute. For DCM, it’s already looking a little worse,
you know? So, there’s no single gene that contributes
the most. We have one that is pretty strong, but then
there is a lot of different other genes. So, you really want to test them all. So, you need an — you need an assay that
can actually interrogate that many in a single test in a given patient. What about allelic heterogeneity? So, as I said, HDM, we have about 11 genes. And we have been testing them in our laboratory
for about a decade now. And what you see here is an analysis of the
result, and it’s asking, “How many variants have been seen in just one proband?” Two, three, four and so forth, along the X-axis. And what’s shown here is that two thirds of
all variants have only been seen once. And there’s a few outliers up here; those
are prevalent but, you know, the vast majority is sort of once, twice, three times. So, that means one needs to really sequence
the entire coding sequence of all these genes for maximum clinical sensitivity. And that’s not just true for cardiomyopathy. If you look at all diagnostic testing performed
in our laboratory — and this was shared with me by Heidi Raine [spelled phonetically] this
slide — we did about 15,000 probands, and the diseases we’re offering are listed down
here. It ranges from cardiomyopathy, hearing loss,
RASopathy, and so on and so forth. It’s actually quite the same, you know? Two thirds or more are seen once, and then
it trails off. And then, this is really the very tricky point
— clinical heterogeneity that has been largely underappreciated in the early days of genetic
testing. And I’ll show you a little bit of why that
was and what the outcome is. So, traditional genetic testing is usually
configured this way; one gene panel for each diagnosis. If you have HCM, you order an HCM test. If you have dilated cardiomyopathy, you go
order a dilated, and so on and so forth. But here’s a case, a real case example, that
we received in our laboratory. This is the proband here with the arrow denoting
an individual with a clinical diagnosis and a family history of dilated cardiomyopathy,
strong family history. The physician ordered what was customary at
the time; a DCM gene panel. And we did detect also what was quite frequent,
a variant of uncertain significance. Now, the variant ended up not segregating,
so we tested all the affected individuals that were available. And it was not present in a few, so unfortunately,
this variant was not the cause of disease, which is quite common. Now, the interesting thing was that about
a year, later this patient was seen again by the physician. And at that point, he revised — and that
was Dr. Lakabel [spelled [phonetically] at the Brigham, by the way — he actually revised
the diagnosis to arrhythmogenic right ventricular cardiomyopathy. That is not uncommon either, because ARVC
has extensive clinical overlap with dilated cardiomyopathy. It’s very difficult to diagnose. Really, the only true way to diagnose it with
certainty is upon biopsy, which is usually not performed, of course. And then, he ordered a second panel on top
of the DCM panel that has already been performed now for this disorder. And this one identified a likely pathogenic
variance, which did segregate. So, what I wanted to — the point I wanted
to make is that this traditional, disease centric testing does not make sense for disorders
with this type of clinical and genetic overlaps. It’s causing a lot of — it’s costly. It’s very time consuming, because between
the first time the first test was ordered and this result I’m showing here, we’re over
a year. So, that was not good for the family and,
you know, and costly. And here is a summary of this concept again. The reason why these older tests were configured
for one disorder only, usually, is that these disorders were typically defined quite narrowly,
based on morphological criteria. And also, these represented the most severe
cases, which is typically how a disorder gets recognized. It’s first, you know, recognized by severe
cases, and then over time, the true spectrum of associative phenotypes is more appreciated. So, this is “the tip of the iceberg” phenomenon. Perfectly understandable why we all configured
our tests this way, but today we do know this overlap with ARVC, but also overlap with HCM. So, really what we’re doing now is offering
— and it’s not just our laboratory. I think the community has moved to multi-disease
gene panel testing because of this overlap. And, you know, we know now that what I showed
you in this isolated case example is true for about three percent of patients with dilated
cardiomyopathy. They come in with a diagnosis of DCM. We end up finding a pathogenic variant in
a gene known to be associated with ARVC. All right. And one final example of this from a different
type of disorder, which is now the RASopathies, which are Noonan spectrum disorders; if you
look at Gene Review, this is what was written in 2012. I actually don’t know if it’s been updated
now. It wasn’t updated as of a year ago, but anyway,
at that time, it says that 80-90 percent of patients with Costello syndrome, one of the
RASopathies, carry a mutation in the HRAS gene. So, that sounds great. So, as a physician reading this, you would
go ahead and order HRAS first, and then, maybe, if negative, you know, do something else. But here’s what we thought. So, this is the abroad referral populations
that we received in our laboratory. And you can see that the minority of patients
actually had a variant in HRAS. The majority had mutations elsewhere in related
pathway genes. So, adhering to this traditional paradigm
would have caused a negative report. And it is unclear how many physicians would
have reflexed to an additional test because again, that is costly. So, to summarize this, these multi-disease
gene panels do often improve a clinical diagnosis. And the reasons are — just to summarize this
phenomenon of phenotypic expansion — as I told you, the original clinical definition
was naturally based on the more severe cases, but then, as a consequence, it ended up being
too narrow as the full range of clinical variability emerged over time. Phenotypic overlap — you know, that is not
uncommon. So, here the disorders present the same and
that can lead to a diagnostic “error.” It’s not really an error, but how is the physician
going to be able to be accurate here? It happens more often, though, as genetic
testing is moving out of specialty clinics to more general genetics care which, you know,
just almost invariably leads to a decrease in detection rate. We’re seeing this more and more because the
cases aren’t precisely diagnosed. And this is now widely recognized by the clinical
and diagnostic community. I’ve had several conversations with physicians
here at Partners Healthcare who are saying, “Yes, this makes so much sense.” We’re actually now beginning to change our
workflow, and this is shown here. So, this is what next gen sequencing has caused,
which is now beginning to be called this sequence first then diagnose workflow. And, you know, not to go through all these,
but this shows the full clinical workload from the patient over diagnosis, and then,
ordering a genetic test and receiving a report. What’s happening is that the up front work
— establishing a clinical diagnosis, and then, using that to order a test — is replaced
by sequencing first, and then, putting the diagnosis at the end; taking the clinical
as well as molecular data into account. And that’s quite exciting, actually. It’s very rewarding to be part of this,
and we do solve cases more than we did before. And as a last background slide, there is definitely
a trend towards genome wide testing, but it doesn’t stop at just sequencing more and more
genes. And this is a little off topic, but I thought
I’d throw it in because it is just breathtaking what’s happening right now. So, we’ve already seen this trend towards
genome-wide testing in other disciplines. You know, in cytogenetics, it was a very,
very early already that that field transitioned to genome-wide copy number arrays; where in
the old days, they were single gene or single analyzed tests like a FISH, a Southern, and
so on. The same thing happened for genotyping tests,
where in the old days, you would look at a mutation, maybe some, and, you know, if you
were lucky, a lot, but not — never so many. And it quickly moved to genome wide SNP chips
and arrays. And now, we have the same thing happening
for NGS — next-gen sequencing. And the reason I’m throwing this all in one
slide is that NGS actualy has the potential of replacing all these tests, because it is
beginning to be for sure possible to genotype, because, you know, you don’t have to interrogate
every position in the sequencing test. You could chose to just look at your SNPs. But it’s also beginning to be feasible to
call copy number alterations, which is a really great argument for using an NGS test. And so, I would say it’s fully expected by
most doctors in the community that this technology will eventually consolidate most genetic testing,
when appropriate. Of course, this isn’t going to be the case
for every disorder. So, I want to move on to talking — and talk
a little bit about which genes should be on the panel. And that gets us to assessing the clinical
validity of variants and genes. The two are connected, so you really can’t
assess the clinical validity of a gene without looking at the variants that have been published
in that gene. And so I want to mention, up front, that this
is a very hot topic in the community, and there are various bodies now that have geared
up to develop standards for assessing clinical validity. And you’re probably very familiar with some
of that. The ACNG and AMP have come out this year — sorry,
last year — with a new guideline for clinical grade variant assessments for Mendelian disorders. And there is also the fairly young consortium
called ClinGen, the clinical genome resource, which is really aimed at uniting medical genetists
and developing approaches to really curate and centralize and share all that data. And its focusing not just on variants, but
also genes; testing which genes are clinically valid in terms of their published evidence. So, let me dive a little deeper into that. Like I said, it is impossible to figure out
the clinical validity of a gene without understanding the published variants, whether or not they
are pathogenic. So, a few words on that; when we assess variants
clinically — I mean, it should actually be the same in research, but in a clinic, we
have a very structured process. We ask, basically, whether the variant affects
the protein or gene function first and then, we ask whether that causes disease. And those two aren’t always linked. And then we classify the variants based on
the available evidence into one of five categories. Five is — these categories that you see here
as the ones recommended by the College of Medical Genetics and EMP. But — and most laboratories begin to adhere
to this. And at the end, we ask one more question. We then take that variant and ask whether
this variant also causes this patient’s disease, because it may be pathogenic, but it may not
be responsible for the patient that I have in front of me. Maybe it’s an adult onset variant, but the
patient I see is a child, so it’s questionable whether this variant is really causing that. So, there is layers in clinical variant assessment. And then, we string together what we see. So, to summarize this again, we go through
the results, annotate and classify these variants into these five categories, and the end result
is that a patient gets the report. And there’s three flavors: positive, negative,
and in between — inconclusive. So, what is classified as likely pathogenic
or pathogenic ends up being a positive report, meaning that we believe that we found the
cause of disease, definitively or likely. So, I want to show you this slide again and
now focus on a different aspect that you might have already picked up on when I showed it
the first time and I just glossed over it. There is this nasty surge in inconclusive
test reports. And we thought, you know, yes, we have a quadruplication
of positive cases, but an even steeper increase in inconclusive. Now this is not good, but it’s worth diving
a little bit deeper to understanding what the causes are, and if all of it is really
bad. So, why more of these inconclusives? There are two main reasons: one is that we
can have a novel variance that has no published evidence, and the variant type on top of it
is of unclear impact. So, this is true for many novel missense variants,
even when the gene is very well established. But the second category is the one that I
want to dive into a little deeper later. It could be a novel variance of any kind,
really, in a gene whose role in disease is not definitively established. And this is really — this is — this is very
prevalent in our community still. But let me just quickly, you know, go over
the first reason: novel variant with no published evidence, and the variant is of unclear impact. So, this is unavoidable. As soon as we start sequencing a gene that
is even definitively and very strongly established with a disease, and we sequence it in its
entirety because there is so much allelic heterogeneity and so many private mutations,
with the good improved diagnosis sensitivity comes some inevitable bad. So, how bad is the bad? That depends on many factors, actually, and
here’s where the physician comes in. It is entirely influenced by the patient’s
ability to deal with uncertainty. It’s also, as I showed you, important to see
whether there’s a family history, because one can turn a variant of uncertain significance
into a pathogenic variant by family studies. And also, the world is moving closer together
now and with many, many large databases being established centralized and interconnected,
we’ve had — we’ve seen an increased ability to solve cases by connecting patients around
the globe. So, it’s not as easy to condemn these BUS’
as you might think. And so my personal opinion on this is, here,
for those disorders with a high degree of allelic heterogeneity, there simply would
never be any progress if one only tested what is already known. You’d be stuck with just five or six common
pathogenic variants, and that’s just it. And here’s just an example, you know, underlining
how you can, you know use family testing — I showed you that before — to make variants
— to move this out of an uncertain significance category. Now, much more important is this: the novel
variant in a gene whose role is not definitely established. And that is a relatively young discipline,
to go into the assessment of gene disease relationships. So, what’s happening is that if a gene — if
the role of a gene is not well understood, you will never be able to interpret a variant
if you don’t understand the role of the gene, for the most part. Traditionally, that’s not been a problem because
the old tests were limited to just a few genes so naturally one would choose those that are
really well established. There was no doubt. But that barrier is gone with NGS. So, all of a sudden there was this possibility
of adding more and more genes. And naturally we all added as many as we could,
only to realize what I just described to you; that “Hey, that isn’t a great thing always
because, we should have actually read the publication a little more deeper and assessed
its validity critically.” Luckily, we’ve all caught on to that, and
now, there is this discipline called gene assessment. And the sad truth is that many published claims
for a gene disease relationship just do not withstand the rigor of clinical grade curation. And it’s not easy to point fingers because
there is the publication pressure everybody has. Journals don’t like to take negative papers. Everybody is trying to hype up their findings. This is all very normal, but it does hurt
you when you use these genes clinically. So, now we actually do this. The Clinical Genome Resource — and I didn’t
mention it; a large NIH funded consortium of many centers — has established guidelines,
and is about to publish them, establishing evidence levels ranging from definitive over
strong, moderate, limited, and so on to none. And then, it has established a rule-based
framework of what evidence is required to make a gene definitively or strongly associated
with disease. So, the pillars of evidence that are used
are, you know, the number of clearly pathogenic variants I reported — and here’s where you
need to understand variant assessments — the number of studies available, the number of
probands with a variant, statistical evidence, other type, case control boards [phonetic],
and then, functional data. All of these things are actually tricky because
one needs to establish rules for what is a valid piece of functional data. But that’s underway. And I wanted to show you an example that I’ve
personally lived through. So, this gene xylene is on our — is on most
laboratory cardiomyopathy panels. And here are the original — sorry — publications. We have one for dilated and one publication
for hypertrophic. And you see the titles, you know? Xylene mutations lead to dilated cardiomyopathy. Mutations in this gene are associated with
HCM. So, that is the statement, but if you look
more closely, in black is what’s in the paper, and then, red is, you know, what you see. And this is a couple of years ago, when you
really look. So, there was good evidence as per the paper,
but when we looked at the variant that they based their claim on — they found a particular
variant — and we found these variants in .3 percent of the population — sorry, in
.7 percent of the population. This was the exome sequencing project. So, no matter what the evidence is there,
this is a red flag. And we’re not saying it doesn’t cause disease,
but it’s not really a slam dunk gene. So, we approached it a little bit more cautiously
today. And even worse for the HCM paper, the two
missing variants that were found by these authors. One of them we’ve already down-classified
as likely benign based on the frequency. And this was pre — this was before The Broden
[spelled phonetically] had released their exact database. I actually don’t know; they might have moved
on to benign now. So, this is what we do clinically. And at the end of the day, the goal is to
develop guidance as to what type of evidence is right for what type of tests. And what you see here is a pyramid with the
different levels of evidence. And naturally, most genes actually live in
this bucket down here — limited or no evidence — then moving up to moderate, strong, and
definitive. And there is no expert — there’s no clear
consensus yet, but most of us include moderate, strong, and definitive in diagnostic panels. But when you go to predictive testing, you
do want to be a little more selective and only use definitive, potentially strongly
associated. But this is, right now, where a lot of activity
happens in the community to actually form expert panels and adjudicate these genes,
and say, “Okay, of the 50 published hypertrophic cardiomyopathy genes, these are the ones that
meet criteria to be included in the diagnostic panel.” And that’s precisely what is happening under
the ClinGen umbrella. Just one example: So, I’m co-chairing a group
for cardiovascular domain — cardiovascular disorders — and we do tackle both. We try to really nail down a framework for
variant curation for this disease. And even more important, we are trying to
establish a recommendation for cardiomyopathy panel testing doing those [unintelligible]. So, now the question is — I mean, no, I would
say — I would like to start by saying in my mind, the utility of these multi gene and
multi disease panels is quite recognized. I don’t think anybody’s debating that. But yes, there’s a higher risk of detecting
the BUSes, and the only negative — that is a negative, but it can be minimized with rigorous
gene selection, as I just explained. How on Earth, though, are we going to keep
up with the increasing rate of gene disease discovery? Disease gene discovery, sorry. As a laboratory director, it’s quite hard
to constantly redevelop, revalidate, and update these gene panels. It simply isn’t sustainable, and in some disease
orders, the knowledge is virtually exploding. So, we need to find a way to keep up. How do we do that? So, this is a big debate in our community
right now. Are we — what’s better? A gene panel or an exome? Why not just — if more genes are better for
some disorders, why not just do all of them? So, this just, like, summarizes the current
landscape. Gene panels are the predominate next generation
sequencing test really focusing, usually, on tens to hundreds of genes. And they come with a lot of perks, you know? We have high coverage. We can usually return a credible result for
every single base in the test, and they’re by and large used for clinically very well-defined
cases. But, there is a big push to go here, because
exome is getting better and, you know, it’s been living in this reach of being used for
complex phenotypes or diagnostic odysseys. But there really isn’t — there’s less and
less difference between an exome and a gene panel, and I’ll show you why. So, a lot of laboratories are trying to figure
out if and when it is appropriate to even move over here. The incentives are very obvious. A large fraction of the gene panels we are
offering are negative. The accuracy rate, at best is often 50 percent
and, you know, that’s just less than optimal. And I’ve shown you before, there’s a growing
appreciation of phenotypic expansions. There’s always been an argument for the hypothesis
retesting. How sure are you that you have the right phenotypes? Well, you don’t, often. And additional tests simply can end up being
more expensive in the end. If you count up all the tests that you end
up ordering, if you do it sequentially, you’re quickly more expensive than an exome. And of course, you have to always be up to
date, dah, dah, dah, and it is also operationally easier to maintain for labs. Barriers — yeah, there are barriers. Cost is one still, though the gap is quickly
closing, so I’d almost disregard this. Incomplete coverage is another frequently
cited barrier, and that is true. Exomes are still not quite as good as the
targeted panels, but a lot of it is a design flaw. The vendors that are offering these have not
done a good job, and so, the community is stepping up to help them with that. There is also an additional risk, and that
is a little more difficult to deal with, which is as we’re rapidly expanding into more and
more genes we’re losing our intimate or any prior knowledge on these tested genes. In the old days — and I’ve launched many
genes over the years — you knew this gene inside and out down to, like, “Oh, here
is an exome that is repetitive or something.” We totally lose that ability now that it’s
possible to overnight test so many genes. But these barriers are really fast disappearing. And I said before, small tests — many small
tests can quickly end up being more expensive. When we look into this a little bit here,
in our ecosystem, this is one example that is representative. It’s a real example; the order test will [unintelligible]
to a sequencing, and that’s blanked out the laboratory. The test consists of 23 genes and included
copy number analysis. The sensitivity, clinical sensitivity was
65 percent. Now we ask the question, if this physician
had ordered an exome, you know, how would it have looked then? And so, in this case, one wouldn’t have needed
additional deltope [spelled phonetically] testing because the exome doesn’t really do
this well. But we factor that in. We would have had 10 more genes at our disposal
because since this laboratory developed their tests, 10 more credible genes game out. And we also looked at our exome quality. It was quite well-covered, so it was legitimate. And the clinical sensitivity would have been
10 percent, actually 15 percent, higher. And the exome turned out to be cheaper, despite
the fact that we wouldn’t have added separate deltope testing. So, this is really how it often goes. The next barrier that is almost gone is the
last of completeness. And this is a slide that I use to show — shown
many times, but I think I need to stop doing it, because it’s no longer a big issue — showing
you what happens to our 51 cardiomyopathy genes that we had in our cardiomyopathy panel
a few years back. If we look at our targeted captured assays,
or the gene panel by itself, it looked like this. In blue, you see everything that is fully
and adequately covered, and there’s a small slice that we would say does not reach the
required coverage; less than one percent. And what we do in our clinical, and many labs
do that, is we use Sanger sequencing to then fill in this slice providing 100 percent coverage. And that, you can see on an exome; the same
51 genes on am exome look much worse. So, only 85 percent are fully covered, 15
percent are not, and that is something that is not possible to fill in by Sanger sequencing. Now, you know, we got together in the community
and helped the vendors develop a better test, which is shown here — an enhanced exome. And you can see the same kind of analysis. And now, the exome derived data was almost
the same as the targeted capture data. So, that’s no longer really an argument to
not do the exome. The most severe barrier, I think, is the educational
gap today. We’re still living up here in this quadrant
— exome sequencing is ordered by experts. If you have, like, a low, medium, high scale
for testing labs and physicians, this is where we are right now. Highly educated, genetics study physicians
order, and highly capable laboratories do the tests. So, we’re doing this now. And it’s just really difficult for a laboratories
to keep up, and also physicians. So, we need to do a lot to educate, but that’s
a separate topic really. So, in my mind, we need to actually redefine
the question we’re asking. So, assuming adequate coverage and assay class
— and I’ve shown you that you think is likely no longer going to be an issue in the near
term future — exome and genome sequencing can be — one has to remember that the way
we’re using exome and genome sequencing can be, can be different. So, everybody thinks sequencing everything
means you have to analyze everything, and that’s not true. So, what we can do is we can genotype. We can run the exome, but we can only look
at known pathogenic positions. We can sequence. We can do panel testing, and if we know the
well-established genes, we can also do all the genes when there’s clinical diagnosis
is not clear, but the family history suggested genetic etiology. And that’s sort of the way exome genome sequencing
is used currently. But price and coverage is really the only
factor that’s gating our ability to use it like a genotyping test or a smaller scale
sequencing test. So, the critical question, really, is how
specific is the patients’ phenotype? That will dictate which set of genes we look
at first, and maybe stop there, you know? And how deep the analysis needs to be. And this is now, you know, really — this
is almost what we’re building right now, right. This is a test we’re about to launch, and
we’re not the only ones in the community that are trying to marry these two worlds, you
know. Can we use an exome? But can we make it behave entirely like a
targeted panel? And that’s exactly what we’re trying to do. So, a traditional disease-focused panel looks
like this. And I said this before; we have 100 percent
coverage using Sanger sequencing to fill in of just, you know, a small number of genes. We typically report deep, like pathogenic
down to variants of uncertain significance, or even likely benign variants. On the other hand, we have exome genome sequencing,
where we also often start with an indication driven gene list. So, if the patient has something that looks
like cardiomyopathy, many laboratories will look specifically at cardiomyopathy genes. But here, typically, the coverage is variable
and, you know, fill-in sequencing is often not done. And reporting is restricted often to pathogenic
and likely pathogenic variants only simply because the scope of the exome is bigger,
particularly when you go down to the next layer, where you start looking at all the
genes and ask, “Is this useful when — you know, to factor in potential phenotypic uncertainty? Is it something that I didn’t expect or, you
know?” And then, there’s also the possibility of
finding things incidentally. What we would like to do is really this: have
a targeted panel that’s doing everything that we’re doing here. So, we’re guaranteeing 100 percent coverage,
and we’re reporting very deep, but we retain the ability and the nice things of an exome
if and when we need it. If this is negative there is — it’s easy
to then say, “Well, let’s go look at the rest. You know, maybe the diagnosis wasn’t accurate. You know, maybe it is worth just looking in
related disorders.” So, that started a new thing. It’s meant to bridge the gap between exome
and panel testing. So, here is a couple of final words on the
importance of standardizing structured gene evaluation. This is also a real example from our laboratory. The goal was to define — to define the contents
of a new indication-driven gene panel, and it happens to be inherited renal disorders. We did a survey of databases. So, it was using ontology driven data base
tools to create a draft list, and it was 279 genes. And then two things: We worked with a clinical
expert and sort of ran this by this expert and said, “Which genes do you think one
should do?” And we then used a ClinGen matrix I showed
you before, which is not rocket science, but forces us to do a very structured clinical
validity assessment. And here’s the result of the 279 genes; the
expert-driven opinion yielded this. So, you know 126 genes were deemed mission
critical, and 22 were nice to have and the rest were, you know, neither; was unimportant,
essentially. When we looked at this with the ClinGen matrix,
we found that not all of them — about a third, actually, a little bit more than a third only
met definitive evidence criteria for a gene disease association. And also, in the rest of the genes that were
not deemed important, we found some that met these criteria. And also some of these genes that were deemed
to be important didn’t meet evidence levels at all. But it just goes to show you that it’s very
important to do this in a structured, rigorous way. And with that, I wanted to summarize. I hope that you can appreciate that multi
gene and multi disease testing can be useful for disorders with clinical and genetic heterogeneity. A genome will soon be cheap enough to be the
first line test for all genetic disorders. And how soon is soon? I don’t know. But that’s where we’re going. And understanding the clinical scenario is
key. The test really becomes an informatics exercise. You can do anything from analyzing just a
few sites. And here, I’m listing an example. You know, provided the genome sequencing is
incredibly cheap, you could just totally run it and ask only, you know, what is present
at the two positions that you need to analyze for achondroplasia. You can do a single gene, if need be. You know, there’s an example — Birt-Hogg-Dubé
syndrome — 90 percent all variants are filament C, so it would make sense to start there first. You can analyze a set of genes. And I took you through HTM or XLM. And curating the validity of gene disease
relationships is probably the most important thing we have to do over the next few years. And with that I’m going to acknowledge just
so many people that have contributed to all these things, and thank you for your patience. And I’m happy to take questions. [end of transcript]