Tip:
Highlight text to annotate it
X
I think it's a little unusual to have two heads of R&D but honestly, I've worked in
this industry for a couple of decades and I've never worked in an environment where
such a huge breadth of disciplines are required in order to develop this technology and I
hope as we dig in a little deeper into the core platform performance that I'll be able
to provide you with some insight into just how sophisticated the work is that's ongoing
that's enabling these innovations that Alan and I have described. So rapid isothermal
template prep is a lot to get around and I don't know if you remember last year and
I think Dr. Wells mentioned it, we called this, had something to do with large volumes
of snow. But for some reason we're not allowed to use
that term. Evidently somebody else owns it. Okay, but what we're really talking about
is an isothermal technology that significantly improves upon a place where we've gotten,
felt a lot of friction from our customers, and that's emulsion PCR. The beauty of this
technology is that it happens in a single tube. So you take your library, some puffer
and a tube that has a lyophilized pellet in it and just mix it altogether. Put it in a
40-degree incubator for 30 minutes and then go on with your ES prep. All right, so from
front to end, this total time is two hours and the great thing is what takes six to nine
hours, depending on which thermal cycling profile you're using on the One Touch-2™, this
is now only 30 minutes. And so we are extremely bullish on this technology.
I'm going to show you that it really is reaching parity with emulsion PCR which some of you
may not know this, probably about 13 years old now. So it's taken a long time to get
rid of this technology. [LAUGH] So this is just a direct comparison of a One Touch-2™
run with the isothermal amplification and you can see by all accounts and all metrics,
they essentially perform identically. When we look across multiple E. coli runs, we're
getting robust and consistent performance. And then when we start to look at AmpliSeq™
human panels, it's actually where you can actually start to see where some of the differences
are and I want to point this out because those of you who are going to be using the early
access will find this out. So of course, as we look against all of the coverage and accuracy
metrics, there's no difference. But where you do see a difference here is roughly 20%
fewer reads compared to the One Touch-2™ runs. And the reason for that is really simply that
we don't have the partitioning that we have in emulsion PCR and we're getting a little
bit of cross-contamination across those amplified particles that are being templated.
And so we actually have quite a few tricks up our sleeve and you know we fully believe
that we're going to bring that performance to parity by the time we launch this product.
One of the things we're super excited about it is the length of libraries that we can
use with this amplification technique. We've used over 600 base pair libraries and the
reason we're not seeing a clean mode now, we're really limited by the sequence and the
accumulated indel errors in these runs now, as opposed to the template itself. All right?
And so we haven't tried this with the Hi-Q enzyme but I fully expect that mode to shift
as we use higher accuracy enzymes and we improve the software. It also works with Proton™. We
actually have used it and we use it, everything that you do normally just ports directly onto
the platform. So those of you who used the early access, of course, we're going to provide
you with all the protocols that we have supporting AmpliSeq™ panels, microbial human genomic,
both Proton™ and PGM™. So this is going to be launched in early access
next month and we actually, you know, we struggled a little bit here because, you know, some
of us want to get this out to you in the next couple months but we really have to be very
careful about thinking about the transition strategy between one type of template prep
and another as well as how we're going to port it on Chef™ or other incarnations of the
technology and so we're just going to continue to develop at this time, but put it in your
hands to begin working with and really, at this stage it works very well. Is Rob Bennett
still here? Rob's the guy, if you have any questions, talk to him. His team's been working
on this for a year. All right. And he's done all the early work.
I'm also going to talk a little bit about the chemistry improvements on Proton 1 (P1™)
before we delve into Proton 2 (P2™). Since we launched the product a year ago, we've
more than doubled the throughput, accuracy, made speed improvements, as well as consistency
improvements in the performance of the platform. And this is basically, remember Proton 1 (PI™) is
the fourth chip we've launched at Ion. PII™ will be the fifth. And I think customers who
have been with us for awhile know that we launched the program and we continuously improve
it and you're pretty much guaranteed to get better products as the technology improves.
Now I wanted to go a little bit into this because you know we've heard you, all right.
Everyone's tired of getting quarterly releases of kits and updates, all right, so we're trying
to calm down, you know, squeeze more together at the same time so there's less transition
for you and really make sure that there is a substantial benefit when you get a new kit
that you're going to have something that you will actually feel the improvement. All right?
And that we're going to be supporting applications that may not have been supported in the previous
iteration. So one of the things that V-3 did, it really
is, when we launched AmpliSeq™ Exome we had to launch the XT kit. What was the XT
kit? XT kit was basically some fancy new ions for your particles that supported the longer
templates that the AmpliSeq™ Exome required. And it also had a better sequencing enzyme
that actually came out of the same Hi-Q screen that Alan described. But we've also got included
in the V-3 kit everything that went into XT plus a 30% reduction in duplicate rate.
This is quite substantial. That brings the duplicate rate down from 30 to 35% down to
15 to 20, depending on your application. So it's basically 15% more sequence data for
every run. We've also reduced the thermal cycling time by 45 minutes, while improving
the GC representation. So not only is it that we sped up the process by 45 minutes, but
now you have much better and more uniform representation across a whole spectrum of
GC categories. And as well, we've done a lot of software improvements, some of which came
out this summer, some of which have come out with Torrent Suite™ 4.0 which have led to improvements
in variant color as well as reductions in total analysis time. So now we're going to
get into it a little bit. You know, we talked about the Chef™. We talked
about how that simplifies a complex workflow. And I know that's going to be a great product.
And I know because all I have to do is talk to our R&D guys and I say, "Would you use
this for your own research?" And they look at me like I'm stupid. Literally. They say,
"What's not to like? I press a button at 7:00, when I go home and there's two chips
ready to put on the PGM™ in the morning." So I hope you'll like that product.
We showed you that we've really removed sort of the remaining weaknesses in our sequencing
accuracy with the Hi-Q enzymes. I think indels are now a thing of the past. We described
an exciting new template chemistry based on isothermal amplification, which is going to
speed up and improve the lengths never before reachable. And we're really, we think it's
going to transform and make it possible to go from library to finished sequence routinely
under eight hours. And I just mentioned P1-V3™, which is... our standard, continuous improvement
and performance that you get with all Ion products. And really this I think P1-V3™ is
important because it makes PI™ enabling for all applications. So when we do product development
at Ion you really are spanning multiple disciplines, you know, from chemistry, polymer chemistry,
surface chemistry. All of that goes into the SNPs themselves.
The amplification protocols that you use to amplify on those SNPs, the biochemistry and
sequencing enzymology that supports the sequencing reaction. The sensors themselves, the loading
of those templated snaps on the sensors, the manufacturing of those sensors, the instrument
itself, the fluidics on that instrument, the hardware, the signal processing at the software
side, the conversion of those signals into base calls, the alignment of those base calls
into variant calls, that whole system, hundreds of scientists working to improve the technology.
But when we think about bringing up a sensor like the PII™, we really are thinking about
four things pretty much. Okay? And I'm going to go into that.
We're thinking about the beads themselves. The micro wells that air, the little structures
that the beads sit in. The sensor, which is obviously it's super important. And the instrument.
So let me give you an update on PII™ and hopefully give you an insight into technology in ways
you've never heard of before. So beads. Ah, one of the benefits of working in a large
corporation like Life Technologies is of course, that you get to work with premiere leaders
in many different disciplines, from enzymology to instrumentation.
In this case I'll talk about Dyna, probably the premiere provider of beads to the bio
technology industry for the last two decades. Believe it or not, the SNPs, we call that
snaps, sorry, the ISPs that you use in your kits are the product of a three-year intense
collaboration with Dyna. We have about 10 people in Norway that we work with to develop
these technologies. And I do want to say, you know we think of them as beads. And one
of the things I want you to walk away with today is that these things, they're really
not beads. Okay, they're kind of like microscopic jellyfish. Okay, they're not solid. They're
very squishy. But what they are and what's so impressive is the process that Dyna has
developed to make beads that are PII™ size 450 nanometers plus or minus 20 nanometers, time
after time after time again, the uniformity of these things is absolutely amazing. And
so we've spent the last six months taking the same technology that was used to make
PI™ beads and just shrinking it and you know honestly, this has probably been one of the
easiest aspects of the project. I mean everything that we've learned on PGM™ and Proton™ just
worked when we got to PII™. And all we had to do was basically make them smaller. We used
the same conjugation. We used the same OneTouches. We used the same loading protocols. And it
works. All right. And this is just an example of some of the
progress that we've made from let's say March to today where we've reduced loading and reduced
clumping while increasing loading to that 80 to 90% range which is standard on Proton-1 (PI™)
So this formulation's done. This work is done. We've made three manufactured lots and it's
ready to kit. Microwells. Probably shouldn't call these
microwells, either. These are really nanowells. All right? And I want to give you some scope
of the scale here. Because if you think about the 3 Series, 3 Series diameter's about three
microns, PI™, 1.3 microns, PII™ here, and you can probably hardly see it, it's .6 microns,
600 nanometers. All right, so we're going from 3 to 1.3 to .6. All right? And let's
put that in perspective. This is a single mitochondrion. Fits in a PGM™ well. This is
a single E. coli which you have to stand on its end to fit into a PI™. All right? This
is a megavirus. I don't know if you've heard about the megaviruses that were identified
in the last like six months. That's how big the beads are that we're loading onto a PII™.
All right? And just to put that in perspective, this is what a high seed cluster looks like
in comparison, to scale. So you can imagine that the ordered array of 660 million wells
with DNA templates that are significantly smaller giving us incredible density advantage,
when we're looking at the throughput potential of these technologies.
So what do you do when you start talking about microwells? Because well you know, it's just
some fancy thing you do in a fab. Should be pretty easy, right? You just build wells and
you know make sure that the walls are 100 nanometers wide and the diameter's consistently
600 nanometers over a wafer that's eight inches wide. You know, so what do we think about?
So we play a lot with the well height, the whole geometry of the height itself. The width.
The cup height. The sensor itself, and I'm going to show you some images of this, that
the sensor, whether it's flat, whether it's a cup, whether it's all throughout the entire
well. And then the materials, there's a lot of materials you can use, both on the wells
and the sensors to optimize the SNR characteristics and the signal characteristics for system
performance. So this is the progress that we've made on
the PII™, again roughly over the last nine months, you see this first incarnation. What we're
looking at here, again, this, just to put it in perspective, 600 nanometers, remember
our beads fit in there. What you're looking at here is the actual sensor. This the pH
meter at the bottom of the chip. And it's connected to what we call a sensor plate,
which is essentially an electrical connector that goes down actually several layers deep
in the chip, to the transistor, which converts the millivolts change in potential that are
responding to the protons, into an electrical signal that we can calibrate. All right? And
so you can see that the way we've been doing this and the reason we've been able to do
all this development work is because we've been able to use the PI™ chip with PII™ wells
on it. Okay, so you can imagine, PI™ has about... 300 million wells, 320 million wells. Sorry,
160 million. Thank you. Manfred's shaking his head at me, 160 million
wells and the PII™ has about 660 million wells. And what we do is we actually build these
PII™ microwells on a PI™ wafer. And so that means basically every other well is actually electrically
connected. And we can run this chip. So that means even though we're building 660 million
wells on the chip, we're using a quarter of them to do the bead development, the templating,
the loading and the data acquisition and the microwell development. All right? So you can
see that then every other well is connected and we've migrated from what is kind of an
ugly, thick walled configuration to this kind of streamlined stealth version here, which
is working really well. So we've been able to spend the last nine months, before we even
have a PII™ wafer we've been able to work out the beads, the bead manufacturing process,
the templating process, the loading process and the microwell. And just to put it as far
as, that's that megavirus I showed you earlier, that was a smudge on the upper right hand
side of the screen, 450 nanometer caps it, fits in a single PII™ well. Okay?
So let's go to the sensor now because underneath those microwells there is a lot of electronics
and it's actually quite impressive. Compared to the PI™, PII™ has four times as many wells,
four times as many transistors. We need twice the data rate to collect that data from the
chip. And um... and we need to optimize our signal noise characteristics so that we can
compensate for the fact that there's actually less DNA going into each of these wells. As
I mentioned in that previous table, we have gone, just like we have with the microwells,
we've gone through multiple designs here to optimize the data transfer rates as well as
the signal noise characteristics, which fortunately you can do in software today. So I want to
put it in perspective about the PII™ chip because you know we think of it as a disposable device.
But it really is quite an impressive thing. The P2 chip is actually a 660 megapixel proton
video camera. All right? And it's transferring about 120 gigabits per second. Now what is
120 gigabits? It sounds like a lot, right? Well, if you take the latest generation CMOS
chip that can do 4k video at 30 frames a second. It's made by a company called Omnividea. It
only generates about 5.6 gigabits per second. If you buy the most expensive hardware that
Cisco has to support routers out in the field, it can go up to about 100 gigabits per second.
All right? So that just puts it a little bit into perspective, just how much data this
chip is generating. To put it in another perspective, you'd need to hook 120 Ethernet cables to
that chip to get that data off, 120. That would be cumbersome, wouldn't it?
And then, or you could watch 36,000 Netflix movies in parallel. In real time. And actually,
to be honest with you, I mean this is, a lot of the reason that PII™ is delayed has been
to accommodate these very high data transfer rates and to ensure that we get stable connections
to the proton when we manufacture the chip. So at this point we have gone through those
design, all those design scenarios. We have at least in the software and you know the
chip guys tell me that the software works. So if the software tells it it will work,
it will work. And actually that's been our experience, certainly with PI™, which worked
right out of the gate. And we're convinced that now this chip is going to be able to
transfer 120 gigs per second. And you know just to put it in, since January we've been
spending all our time really going through those design iterations, doing that very sophisticated
modeling of the chip performance, making sure that all the links work and that the electronics
are maximally quiet. And then you do this whole thing where you do logic verification
testing and design rules. Basically what you're doing is you're making sure that there's no
interferences in the electronics and then when you actually tape out, tape out is just
saying you're actually making the masks for the 30 different layers that comprise this
chip. And then once that database is transferred to the fab, you can start building them. And
so that's where we're at today. We're really just at that transition right
now between tape out and wait for fabrication. This process here takes about six weeks. And
then we get to integrate all that stuff I described to you. The beads we've already
done. The microwells, the templating and all that stuff and get it to work. So the Proton™,
there's not much to talk about. Quite honestly, when we delivered the Proton™ to you last year,
we actually took a pause and said we want this machine to be PII™ ready. The electronics
are already in place. In fact, it has substantially more capacity than it needs to support the
PII™. We've already shown that it's able to sustain that 160, 130 gigs per second. So
um... we're satisfied that the instrument's going to work. What you may not know is we
actually do a lot of runs in Ion. I think we probably have probably over 100 Protons™
across three sites. And we've done over 25,000 PI™ runs and just
for the PII™ project alone we've done over 2500. And you know the data, the feedback we're
getting from the field that, not just in our own hands in house, we're probably the biggest
user of Proton™ chips in the world at this point, is hopefully not for long. Um... this
thing's rock solid. We're seeing mean times around 11 months, out in the field. We've
also again made some substantial software improvements and in these analysis times what
you can see here is that from the time we launched this instrument last year we've significantly
reduced the total run time. What's that all about? That's just not about making faster
data off; it's also about being ready for PII™. Okay, that means that this machine is
going to be able to handle the volume of data that we need to handle the four times more
sequence coming off a PII™ chip. So, in summary, you know I think what I want
to convey to you is a sense of confidence. All right. We can make the beads. They work,
they template, they load. We can make the microwells. We can hook up that chip to the
machine and the instrument. And we have a sensor that we're convinced is going to perform
well and you know we're just as eager as you are to get it into our hands and into your
hands and so that's what we're working on. Just to give you Mike to update, I did update,
I don't know, but um, this just gives you a sense of the rate at which we're increasing,
we're actually going into an exponential phase of improvement. This is all of that bead and
microwell work converging. Since March you know we're getting routinely now over 30 gig
runs. We actually have generated some very interesting
data. So um, what you see here actually, I've got to cheat, MAQC standards that have been
run in triplicate and you can look at the R values here. We're getting extremely high
replication rates. This is on that PII™ prototype chip that we've been doing all the development
work on. This is even before we get the real product in our hands we're seeing tremendous
reproducibility that's as good as any platform we've seen including Proton™ and PGM™. Proton
1 (PI™). If we take those replicates and we do a differential
expression of profiling analysis, and we use 907 gold standard Taqman™ assays, um, you see
we're getting R values of .94. So that means not only the reproducibility between replicates
but also in the experimental comparisons and again, this is on a PII™ prototype. So we fully
expect by the time the PII™ comes out and we're continually getting 300 million reads, you
should be able to multiplex anywhere from five to ten samples in a single run to get
your expression data or any coming application for that matter.
So, I think I'm two minutes over so I'm just going to tie this up quickly so we can open
it up and hopefully you'll have some questions for us. We've really been focusing again globally
on the ease of use, the workflow with the Ion Chef™ system. I think we've addressed any
remaining indel questions that may be in the platform with this Hi-Q chemistry. Um, throughput
of course is coming with PII™. And we're going to make our system even faster with our isothermal
amplification technology. Thank you very much.