European Society of Human Genetics Conference (ESHG)

*This transcript was generated using an AI-based transcription tool and may contain errors. It reflects the spoken content of a live webinar and has been lightly edited for clarity.

Welcome everybody. So, a couple of things. We did an introduction back in the February webinar. There's quite a bit more detail in that that we'll cover here in terms of technical detail, especially John Mannion, who's on the right there. He went into a lot more data path kind of information there. So check that out. If you haven't seen it, there's additional information. 

We also published the seminal preprint at the same time, and that covers the chemistry up through 2020. It's a good read. If you're into the chemistry, I think it's a good thing to get more background on, as well as AGBT, where we kind of did the launch introduction of technology, where Hartwig and Brode and Sean presented, and he'll be presenting again here, covered a lot of good stuff there as well. 

So, a lot of good things to look at to get background. So in that sense, I'll just really touch on the technology overview a lot faster this time, kind of try and go through it pretty quick. So I want to be able to focus on some of the data we're going to show today. So we'll do an update on the SBX duplex and the fast sequencing approaches. 

So we've actually improved the data from even two or three months ago, and I think that'll be the theme as we continue to advance the technology.

We'll show some of our first-ever FFPET data that we generated using our duplex approach, as well as MRD, and then finish off with some simplex applications of RNA and DNA. So that will be kind of the agenda for today.

So when we conceived of the idea, I've said this many times, it was born out of flexibility. We really wanted to do a single-molecule approach that could flexibly sequence. 

For me, if I wanted to sequence for four minutes, I'd want to be able to do that. I wouldn't want to wait around a couple days. I'm impatient. Or be able to sequence for several hours to get massive throughput. I wanted a technology that can do that, and so that was the thinking for our approach to doing single-molecule sequencing. 

Now, for every approach, you have to be thinking about performance, you have to be thinking about scaling into the future, and so we'll cover on accuracy, and we'll talk about what we're doing, you know, update some of the F1 scores that we have using the genome in a bottle samples, quite, you know, they've improved quite a bit even in just two or three months. 

Throughput, still very similar, 500 million bases per second throughput, and then we'll go into the distinction between the simplex reads and the duplex reads, simplex being really longer, higher throughput, Q20 accuracy, whereas the duplex is the higher accuracy reads. 

We'll update on some urgent samples, basically this SBX fast approach, where we go from sample to VCF now in less than five hours, and Sean will cover a lot of that in some of his presentation as well. And then, of course,

At the end of the day, cost efficiency. People ask this question a lot about what's the cost going to be. And, of course, this is top of the mind for us. We have to be efficient, and the technology was built for that. We think about efficiencies in terms of measurement, the quality of the data as it comes out, so you can efficiently analyze the data coming out. 

All this has to come together if you're going to be able to drive your costs down, and that's critical. And, of course, we want to be able to give people the kind of technology that they want moving forward, and that's key. Otherwise, I think, what's the point? So cost efficiency is a critical element. So what is this? What is our approach to sequencing? It's the coming together of the stratogenomics sequencing by expansion chemistry with the Genia high-throughput sensor array.

Okay, the way that looks, I don't know if you've gone to our booth, you can see that we have both of the systems there. The SPX chemistry happens on the synthesis instrument, and then the sequencing on the sequencing instrument. Two separate approaches there, or systems there. And we actually think that's a more efficient way to run for the time being. Down in the future, you know, I would imagine we would have a system that could integrate both workflows for specific applications. But right now, flexibility, we think, is maximized by keeping them separate.

So the general approach was simple: don't sequence DNA. I think if you want to do single molecule for us, was rescale the problem. It came down to signal-to-noise for us. How to adjust signal-to-noise of the measurement was the problem, so we created this new molecule called an expandomer, a surrogate for the DNA molecule, where we could rescale that measurement problem. 

At the time, nanopore really wasn't even still a thing in 2007, but over time, it seemed obvious to us that we wanted to translate the technology to a nanopore sequencing approach. And so, you know, very difficult protein molecular engineering to be able to achieve this, but when you get on the other side of it, it does open up a whole lot of possibilities on the measurement side. So how do you do this? I think the key, one of the key elements, is the expandable nucleotide or XNTP.

When we conceived the idea, it was really about putting a reporter code, a reporter tether, and attaching it at two places on a nucleotide, alpha phosphate for one, and off the heterocycle for a second, and then separate that with a cleavable bond. 

Okay, so that was kind of the beginning. Over time, as we directed more towards the nanopore measurement, we knew we wanted to have something that could control translocation as it moves through the nanopore. And so we put that translocation control element on there.

And then of course, over time, we learned that, you know, we really need to enhance the incorporation of these massive nucleotides, and so we built in enhancers. I can remember in 2014 going to a meeting at Roche, this is when we were still partnering and working together, showing a gel with an extension of a single XNTP, and I was really happy to show this. 

I mean, this was like great. Okay, we've made the XNTP. We can get a single extension. That's when we kind of realized, yeah, we need to enhance the incorporation of these nucleotides. And so we engineered those into the structure of the nucleotide, as well as many other things, and that's the basic, the foundation was really a molecular engineering exercise to make these modified nucleotides. 

Many other things came as well. I'm not going to cover all of it here, but there are a lot of other elements that you can get in some of the other talks that we've had and certainly our paper.

So the basic idea of the synthesis instrument is liquid handling reaction control system to synthesize the expandomer on this chip that we show in the top left panel there in the green box, as well as a bunch of reagents to basically support the biochemical synthesis and conversion of the DNA template into an expandomer. 

So it's pretty intuitive. You have an oligo on a surface, hybridize a template, throw in a cocktail of expandomer synthesis reagents to copy, do a template-dependent copying of the template. Cleave at that amidate linkage I showed you earlier to expand, and use an orthogonal cleavage, photocleavage, to release the expandomer from the support, releasing a pretty clean molecule, because you want to be able to measure this in a nanopore, and you don't want a lot of noise there.

Another key element was, as I mentioned, was translocation control. You can see that little green triangle, a green arrow pointing to the triangle there holding an expandomer in place. It's on the barrel side of the nanopore embedded in a bilayer, so you can see it kind of holding it in place so you can get good statistical sampling at a base voltage, so you can get that signal that you see with that green arrow showing where the G base is.

Then with a very quick microsecond level pulse you can advance the expand emmer to the next level and that shows now a C code coming in and so on and so forth and with this approach we couldn't we can modulate the throughput at a very controlled deterministic way and that one adds efficiency but adds quality to the measurement this is all about efficiency of measurement at the end of the day and so this is a big deal when we first made the company, it's actually many years later, we drew this up. 

That's actually a drawing of what we wanted to do. And, you know, this is the dream of what we tried to do. And a mere five or six or seven years later, this was actually a real expandomer trace once we figured out how all the elements that had to come together to make these expandomers. 

This is a result on a single-pore axopatch system that we did at Stratogenomics. But you get an idea of that signal resolution we were able to achieve. So we'd solve that part of the problem.

You get an idea of what a homopolymer looks like, three C bases in a row, you see the resolution there, and then of course, the level histograms, how nicely separated they are. I mean, this is critical for single-molecule measurement. If you're going to have accurate sequencing, you can't have overlapping histograms. And so that was, so we'd achieved this, I look at this as kind of like the baby picture of the expandomers. 

So we look at this, you know, fondly, but then we had to transition. So now moving it into a much larger array, now instead of one nanopore, we're on an 8 million array where we combine microwells, electrodes, and detection circuits to do the analog to digital conversion. And you can see in the right hand there, you see a single microwell with a bilayer, that orangish color, a pore, and an expand, and we're threading through. So now 8 million times the throughput.

So the sequencing instrument, again, fluid handling, but of course, all the data coming off the system. The sensor module is a key part that contains the chip, the 8 million chip on it, and it allows for the fluidic control of all the reagents to go through. And a process, a very simplified version of the process, on the right, there is the sequencing workflow where you set up the bilayer, do the pore insertion, throw on your expandable reagent to do your sequencing, clean the system, and reuse. 

This is key. I mean, as you can imagine, this is a pretty sophisticated device. And so being able to reuse is pretty critical to get the cost profile down. And so this is something that I think really is very important for the chemistry, that you literally go through that workflow, that cyclical workflow, over and over and over again, and I think that really helps us drive the cost down significantly.

And just a picture of the pores. So remember, it's an 8 million array. So time zero, typically we would get into 7 plus million pores that are actually sequence quality pores, which is an astounding number, actually, when you think about Poisson. And then over time, as you can imagine, you know, pores will clog, or you'll have a bilayer break, and so you can see what it looks like over, say, a four-hour run. 

You still have five million functioning sequencing pores. But all the throughput I'm showing for the longer runs would have gone through that type of fall-off of pores. And that's also another opportunity for us to improve throughput into the future.

 Okay, this is now what a single pore would look like on the 8 million array, and you can see that arrow pointing to open channel, which is no expandomer flowing through, and you get an idea of the signal separation of many hundreds of expandomers, maybe even thousands of expandomers going through.

But when you zoom in, you can see what one expandable molecule looks like. That's probably a 900- to 1000-base sequence, and if you zoom in even further, you get an idea of the signal resolution for the translated XP base calls there. So still great signal resolution on the 8 million array.

And again, flexibility is key. We want you to be able to do one expandomer or four expandomer pools, whatever the multiplex level you want. In fact, that's the synthesizer does up to four samples at a time. And if you have, you know, four minutes to four hours of sequencing you need, we want to be able to allow people to sequence through the whole flexibility throughput spectrum there on the left here. So that's the goal, has been the goal of the technology from the very beginning.

Okay, so moving on to duplex sequencing. We've covered this several times up to this point, but it's basically the structure there, the library structure there shows a Y adapter and a hairpin so that you can actually generate an expandomer and do intramolecular consensus. This allows us to get very good accuracy with the single molecule technology.

And the applications and some of the things we'll talk about today is MRD, as well as some of the rapid whole genome sequencing and FFPET. So, the MRD and FFPET, this will be the first time we've talked about it. This is brand new data, and we haven't really started working on it in February. So this is all new stuff that we've been working on since the last main presentation we did.

So the workflow for SBXD is pretty straightforward. You fragment a library to whatever links that you want and do this ordered ligation to put a Y adapter hairpin on. I show the amounts that we're typically using for genomic DNA and cell-free DNA.

We then go through and do a linear amplification, which is one of the unique things about what we do. Very much helps us with our accuracy, being able to do linear amplification with homopolymer regions. Other than that, it's pretty normal synthesis that falls along the lines of the workflow I showed earlier.

A little bit about read structure there. So on the left you have some raw read results, what they look like. On the right shows this partial duplex read, shows pictorially what it actually is. And in terms of these partial reads, they're exactly what it says. Some of it's duplex, some of it's simplex. We use all of that information.

We don't want to throw it away. The duplex spaces are what we use to actually give us our concordant accuracy or reads that we get. Simplex spaces can actually be used to help us with mapping. So we don't throw it away. We use it, but we only use the duplex spaces for our accuracy determination, and then below we see the full-length reads, which you get an idea of what those look like. 

On the bottom, you can see the hairpin adapter sequence in the middle. So you get an idea of the length histogram that we typically see for expandable links.

Okay, a little bit of update on the data. So we showed this previously. These are genome-in-a-bottle samples, one-hour sequencing run, 5 billion duplex reads in one hour of sequencing. You get an idea of the median coverage that we see and the mean insert length. 

Those are the lengths of the full-length duplex reads that we're showing there. So around 240 to 250, median coverage, high 30s to low 40s. and then the F1 scores for SMBs and indels are shown in the table there. And I'll point out that we have both our GATK Roche machine learning approach on the left, as well as the deep variant from Google. Great collaboration there. Really awesome team to work with. 

Their numbers, you can see, are a little bit better. But you wanted to acknowledge that relationship there.

So, as we said, accuracy since the last time has improved a bit. Q40, average duplex base quality, and 99.9% coverage of greater than 10x coverage for the genome. So all of those numbers are improved from just two or three months before.

And this kind of shows in terms of F1 scores, the dark blue being where we are currently and the light blue being where we were in February, which is already really good numbers for an amplified library, and so we're continuing to push on that, and we'll continue to do so.

And this just shows the results if you downsample the 30x. Typically, that's usually one way of looking at it. They really don't change that much. Maybe a 0.1 or 0.15 drop in F1 scores going to 30x. Still very good results.

Okay, looking at the homopolymer links, you can see that for all homopolymer links, less than 17 base pairs, we have F1 scores greater than 99% using either pipeline, and that's on the left is the Roche pipeline, on the right is deep barrier. Very, very good results in terms of homopolymer. Coverage on the left, you have the histograms of both all reads and just a full duplex reads on the left.

And you can see in terms of the coverage across the GC spectrum, the low GC, you know, look at that, still really happy with that little work we can do there on that, but still very happy with what we're seeing of the low GC. High GC is looking pretty great. So we'll always continue to improve. Next up on the list will probably be to lift up the lower GC on this, but still very, very good data that we get out of this.

Okay, and then accuracy across the links, varying links of expandimers. This kind of shows just no drop-off in FRED score as you go through the number of links. So like, for example, the 300 base insert there, you see a pretty flat line with FRED scores of 40 across the entire length of 300-mers, and there was 350,000 of them used to plot this data here. So no drop-off in accuracy across length.

Okay, SBX fast, very similar to SBXD, only we don't do the linear amplification step. We put more sample in, in this case we put two micrograms in, but then we don't have an amplification step here. Two micrograms is really driven by the level of plex that we have. If you're not going to plex deeply, two micrograms is what we go with. But if you wanted to do higher multiplexing, you can get away with probably less DNA into the future. 

Okay, so amplification-free workflow, and the results here, we show something similar before. Very, very good. F1 scores again for these. We ran these as trios because that's going to overlap more with what Sean's going to talk about for the direction of the SBX fast workflow is to run solos or trios. So we ran these as trios, so we actually have a lot higher coverage here, but still very good results.

The difference between February and May, still, you know, big jump. And a lot of these improvements were, they were chemistry, they were data algorithm analysis, the range of things. There just wasn't one thing that led to these improvements. And so we'll keep working along those lines for all the chemistries and all the approaches we're working on right now. Same thing when you downsample 30x, still very, very good F1 scores.

Homopolymer now, instead of 17 bases, we have F1 scores greater than 99% all the way up to 21 bases, so even better, and that's the difference between doing amplification or not. But still, for the amplified results, very good homopolymer accuracies.

Histogram, very similar to what we showed before. Same thing with read depth and coverage. Very, very good there. Continuing to work on that as well, but I think so far, I think very, very good coverage across the GC spectrum. Opportunities always, though, to improve.

So in comparing February to May, we can see the read length differences between SBXD and FAST going from 230 to 248, both increased there. The improvement in FRED scores to Q40 and then coverage improvements. So across the board, we've improved the chemistry and the detection.

This is a fun one. This is something that we've been actively working on since February is SBX fast time trials in a solo sample. How fast can we process a human genome? And so the workflow we showed above here, going from library prep to expand them or synthesis sequencing, and then analysis through variant calling.

And so those are the time windows that we ran. So now in February, we were at six hours and 25 minutes. We're now at four hours and 23 minutes for library prep through VCF, and there's still more room to improve on this, so we'll keep working on this. But our goal originally was to be able to do trios within a normal workday. I think we're going to be well within that. Sean will cover some of that in his talk.

So we were able to detect the variant. In this case, it was a Coriel sample with propionic acidemia. Able to positively detect that, 30x coverage. Yeah, in just a little over 4 hours and 23 minutes. This did not include the actual prep of DNA prep. So add another 20 minutes to that, we'd be at about 4 hours and 43 minutes for total sequencing from blood sample all the way through variant calling. So pretty exciting.

It's fun to work on that and fun to continue to work on that. And in fact, we just announced this last week, I believe, yeah, that we were doing an expanded collaboration with Broad. So we're going to continue to do a lot of this, and I don't know if Sean can maybe mention that in his talk.

So one of the other key things for these solo SBX fast runs was being able to do the analysis and read mapping and alignment basically concurrent with the sequencing itself. This comes down to what I was mentioning before, the efficiency of measurement and the efficiencies that you have to be able to do throughout every step in the process like this if you're going to get your cost down. 

Well, in this case, we were able to actually do DMUX intramolecular consensus mapping and alignment concurrent and that involved, you know, a lot of improvements in the implementation of literally every accelerated algorithm in the pipeline that we're currently using. But in doing that, we're essentially able to get rid of hours of time of sequencing.

So this is a big part of it and wanted to acknowledge that and great relationships that we have with NVIDIA as well as working on the Parabricks libraries with the team. But a big part of this, along with some of the chemistry adjustments, of course, that we've been making to get to that really fast genome sequencing time. 

Okay, moving on to some new stuff. So FFPET.

So very similar-looking workflow, SPXD-like workflow. In this case, we actually did some DNA repair steps. Before, we found we got better yield and actually better accuracy by going through a DNA repair step prior to doing the duplex library preparation. Again, linear amplification as well we used for the process. But other than that, it was pretty much the same as the SPXD protocol we normally run.

So in terms of samples, we did 18 matched normal tumor samples that we ran through with the range of Q ratios that were determined by the line assay. So you get an idea of the kind of the ranges of quality of the samples that we were using here. Normally, we don't do these type of comparisons, but we needed some frame of reference. 

So we just figured we'd list, you know, exactly kind of what we ran here. So in comparing to aluminum sequencing, we use the NEBNext repair kit, 100 nanograms for each sample going in. For SBX, we did the SBXD protocol. For Illumina, we used the Kappa EvoPlus V2 library prep. 

For Illumina, they did the NovaSeq 6000 S4-300 cycle kit. All of the data that we actually included in the summary here were greater than 70x coverage were included for comparison. And for Illumina, we used the Dragon 4.3 for analysis.

So looking at the data, the accuracy for the, basically SBX shows lower SNV indel error rates higher long homopolymer accuracy when compared to Lumina. I mean, when you look at the previous data, it kind of makes sense. We're doing really well with the long homopolymer. On the left, you can see the FRED scores are looking very good for SPX, relatively speaking. Error rates also looking good. And then you see that nice profile, especially looking out at homopolymer links of 8 all the way through 20, how well they compare with this, again, amplified protocol. 

With a lot of FFP samples, it kind of may make sense to be able to use an amplified protocol, but in this case, looking quite good. And this carries on through the rest of the analysis we're doing for FFP. So similar on coverage, SBX showed better coverage uniformity in the higher GC regions that I show to the right here. Overall, though, pretty comparable there, looking good. But that higher GC, I think, is definitely, for us, a real difference.

And then evaluating some of the not difficult regions, the NIST not difficult regions of the genome, using similar false positive rates for both SBX and Illumina as determined by normal-normal analysis. And so we use 0.08 false positive rate for the...

 SMVs and 0.05 for the indels and again you can see the differences between the two systems of two platforms there so SPX generally showed about 10% better recall compared to Lumina in this comparison so again first time we've ever done this so not like we've run it thousands of times but we're excited to show and this is definitely be a focus of additional work as we move through the summer and fall

Just to give an idea, the allele frequency distributions, they were pretty much overlapping, very similar between both technologies. So looking at now some of the more challenging regions of the genome, so SMB calling there, reasonable concordance between both SBX and Illumina. You do see SBX achieving higher sensitivity at the similar false positive rates, again, due to some of the data we showed you before in terms of the lower SMB error rates and the better uniformity in high GC.

So you can look at basically the difference between the blue and the orange here, where SBX only has a much higher amount relative to some of the aluminum. Okay, similar here with indels, even more dramatic here with indels in terms of the difference between the two platforms, but overall, you know, pretty concordant, pretty good, and certainly with SBX showing better sensitivity.

So for copy number variants, this is about 10 of the samples shown, so it makes it a little bit noisy here, so we zoomed in on one copy number variant example from one sample to get an idea of the level of concordance there, but I think pretty good concordance. We're not trying to say who's right or wrong in either of the discordant, but I think overall, very high level of concordance.

And this is a zoom in on a single sample, again, so you can get an idea of both the arm level and focal event differences between the two technologies and very good concordance there as well. So overall, I think pretty sound, pretty solid data. So MRD, moving on quickly to MRD.

So we started off with four nanograms of DNA. These are some samples that we have pretty limited. We wanted to challenge ourselves here. Again, SBXD protocol that used linear amplification and just brought it through our standard SBXD workflow.

Okay, and as far as the sample cohort, we use 15 cancer samples at various stages, 15 healthy donors, all four nanograms each for each one of them. This is kind of the breakdown of the different samples with tumor fraction there.

Okay, and a little bit about the workflow here. On the left side, you see that we go from our cell-free DNA to our consensus reads that we talked about before, getting around a SNP accuracy of Q44. We then do the alignment to reference and do super consensus collapsing to get to a Q46 accuracy, like level of accuracy, and then additional read filtering, trimming of the ends just a little bit, and we got to a Q50 type accuracy. 

And from that, we can get this estimated tumor fraction number, and so with the comparison for all 15 of the samples, we were able to see positive MRD for all 15 of the samples we looked at and would assess that we were at the 1 to 10 to the minus 6 estimated tumor fraction for these and get an idea of the breakdown between the tissue, different tissue samples and the stages of the cancer for each of the samples. 

But we're really happy to see this again first time that we run this type of experiment. So these are all going to be kind of firsts here. And this will continue, I think, as we get through the year, as we do more firsts with different applications and expand on this with some of our early access partners. So pivoting now to SBX Simplex.

So SBXS, as we call it, are going to be reads less than 500. It's a little definitional thing here. SBXSL are going to be reads that are longer than 500. We kind of put that as the limiter here. So we're going to talk a little bit about the both, but I think mostly about the SL applications.

So on the right, it's pretty much as we said, it's not that duplex read structure, it's that simplex single read through, higher throughput with the potential to do longer reads. And this is helpful for, say, RNA isoform analysis, which is something that we're not going to talk about here. 

We did last time Aziz El-Khafaji talked about it at AGBT, and I'm sure in the coming months we'll talk a lot more about some of that work. But you can imagine doing probe-based, like 10xFlex applications for this approach, and as well as some of the RNA isoform stuff. And so I'll just show a summary here of kind of a throughput summary that I showed before on the left. 

You get an idea of the output of 10xFlex reads in a one-hour sequencing run of about 14 billion reads, and that's 14 billion reads that are what we call cell ranger valid reads, which are valid barcode, high-confidence reads.

So it was actually over 20 billion reads total, but we're really strict on how many we actually brought into CellRanger. And so about 14 billion reads in one hour of sequencing. On the right is a 10x 5' mRNA run. You get the extreme other side of the equation here, where you look at, for all reads, about 4 billion in an hour. 5%.

500 base pair average read length, so you can get the idea of where we're going in terms of directionality, in terms of the length and yield for the longer read side, and this has actually improved quite a bit since then. Okay, so that's all I'll cover on RNA, but we're actually going to go do a lot more in the coming months on the RNA applications.

So on -- but expanding on the simplex side for phasing, for genomic DNA phasing applications, was something that we were interested in doing for a long time. Now we're taking this as an opportunity to kind of keep pushing on that workflow. Looks pretty similar. 

Y adapter ligation, pretty straightforward, but now using the linear amplification, which again has the advantage of being able to drop straight into expandomer synthesis. It's already strand-enriched. That's our first step, or last step in library preparation is that strand enrichment. These can drop straight in to SBX directly. Very good quality. Again, homopolymer advantage there as well. So this is the workflow we've been working on.

Okay, on the left, when we're thinking about variable number tandem repeats, variant calling, on the left kind of tells the picture. If you're expanding to VNTRs of 200 or 800 in length, obviously, they're just not going to be captured, not going to be able to resolve those. 

On the right gives you an idea of what you can do with 1,000 more SBX reads, especially if you can get a billion of them in one hour of sequencing at 1,000 base length, and the proportion of ENTRs that you can actually detect there. And this is a specific result from HG002, genotyping concordance kind of benchmarking run. 

On the right actually gives the numbers that we got. Again, this is a one-hour sequencing run, billion reads, average read length almost 1,000, so almost a terabase of sequence of just those length.

In one hour sequencing, so there's a lot of power there obviously we're going to keep pushing the length up beyond the thousand go to 1500 be able to get better phasing as you get longer, but this is all pretty new stuff that we've just only started working on and on the left, you can get the idea of the alleles that cannot be phased with shorter read, and that's SBXD, that's Illumina, that's any of the shorter read approaches. 

You wouldn't be able to phase those various alleles shown there. But on the right, you see SBXSL that would be able to, the CNT alleles would be detected in phase while the G allele is out of phase. So it just kind of frames the picture there.

Again, another way of looking at it, in terms of F1 score, the dark green being the 800 base and greater reads, what that looks like across the maximum distance between consecutive variant number there. So quite a dramatic difference between the SBXSL and what would be the shorter SBXD reads or Lumina or any other technologies on the shorter read side.

Okay, and this is just one last thing on this, which was a pathogenic compound heterozygous variant from a sample that was affected with hyperphenylalanemia, easier to say.

And we were able to correctly detect with SBXSL, but not with Illumina or SBXD for that matter, be able to actually correct this heterozygous variant in this particular sample. Again, similar read throughput here, a billion reads in one hour of sequencing. Okay, so summary really quick. So we covered the SBX fast. We're now sub-five hours from sample all the way through sequencing.

We've now started to work through FFPE, MRD workflows, and are expanding our work on longer reads, Simplex, and we'll continue to do all of these. A lot of these are firsts, which is why it's exciting to come and talk about this stuff. And Jagdeesh is right back there. His team, he's smiling. He's the guy right there.

Yeah, wave your hand, Jagdeesh. It's okay. Yeah, there you go. Anyway, his team's doing a lot of the work on the chemistry side to pull all this stuff together. And it's, you know, I think throughout the year, we're going to continue to do a lot more of these kind of application developments. So in terms of the future, there are more buttons to push on the technology than we can actually push at any given time to optimize the measurement. So there's a lot to do.

We will continue to relentlessly do that into the future. That's the nature of this approach. This technology is still very early in our understanding of how to utilize SBX. And so we'll continue to hit on all of these fundamental optimizations here as we embark on the next step of SBX as we get towards launching in 26.

So a lot of things there, and I know this will probably could come up in some of the questions, but next steps for us will be showing target enrichment, some methylation work, and a range of other different application areas that are very interesting to us. So continuing to work on this and excited for the next steps, and I thank you.

*This transcript was generated using an AI-based transcription tool and may contain errors. It reflects the spoken content of a live webinar and has been lightly edited for clarity.

Thanks for the intro. Me too. Thanks for also walking through my whole CV. That was great. So I'm going to talk to you guys about pretty much adding on to what Mark talked about. We've been focusing a lot at Broad on trying to have not just proof of principle that this technology can work, but how do we actually start thinking about this in practice, especially in the clinical space?

So, Broad Clinical Labs, we're a part of the Broad Institute, also known as the genomics platform within the Broad Institute, really large sequencing center in the U.S. based in Cambridge, Massachusetts. We serve quite a lot of different groups that span across different types of partners. So, we have some groups that are like really trying to push the boundaries of genomics. 

So, how do you do, identify new types of events, add new technologies, really like the people who are kind of creating the groundbreaking research. We then also have a lot of groups that are doing like resource building, so a lot of the tools that a lot of people in this room end up using actually come out of the Broad Institute. 

And also we're a part of like large consortia, so we generate a lot of data, and we make a lot of that data publicly available. And then the last group are really clinical translators. So this is kind of a newer arm of Broad Institute. And that's where our group primarily focuses. And how do we take this technology and these large data sets and our scale and actually apply them to clinical problems?

So, here's a couple of the different partners that we have in the different groups, and some of these may be familiar to you. Most of them are based in the US, so recognize them that this is an international community, so some of them may not be familiar. So we are a large-scale sequencing center, so this is just looking at our current fleet. So we have a large setup of NovaSeq X+'s that are pretty much running non-stop around the clock.

We have several of the Ultima systems and then 10 PacBio Revios. On the right-hand side, you can see some of our midplex and smaller sequencers as well. So we're pretty much platform agnostic, and we actually really like to have the opportunity to engage with vendors as they're first working on their technology. So we tend to be like a proving ground for a lot of new technologies, which is a really great opportunity.

And that's actually kind of how the conversation with Roche started, I guess about like pretty much a year and a couple of months ago was when the first conversations were taking place. And in those conversations, the Roche team, like Mark kind of went through a lot of the initial data, had some pretty impressive claims on the ability of the sequencer to have high accuracy, high speed, and high scale. 

So basically, having the ability to have a lot of runs, smaller runs and smaller batch sizes leads to actually a huge amount of scale that could come out of one sequencer. The Nanopore and Expandimer sequencer has a lot of potential with accuracy, as Mark talked about. 

It was great to hear about all this, and so we were kind of thinking about what the applications are for this technology. So what are those different aspects and different things that the sequencer is really good and the SBX chemistry is really good for? What are the best applications for this? And so immediately started thinking about rapid genome in the neonatal intensive care unit.

This is a figure from a publication that came out of Stephen Kingsmore's group. But in general, the ability to have early diagnostic testing by rapid whole genome ends up saving a significant amount of burden on the patient, the healthcare system, and the hospitals. 

Around 60% of the patients end up with positive findings. So the NICU, even patients that you don't expect to have monogenic genetic conditions, once you actually start testing you end up seeing that there's a large percentage of these patients that have it, there's not always as we all know, like for monogenic conditions, there's not always therapeutic intervention. 

Now there's more and more that are hitting the market but even just having an answer helps a lot with the management and management decisions both for like the neonatal team and for the parents of this child who's severely sick, and so being able to get an answer fast is critical. 

And people have been doing this, so rapid NICU genome testing has been around since, I don't know, probably around 2015, 2014. And RADES has really been kind of championing this in the U.S. And then...

And other groups have as well, and now there are commercial offerings for rapid testing. However, that rapid testing, it still takes a while. So rapid could be anywhere from a week. It could be four days. Now I've seen an advertisement for like an ultra-rapid at 72 hours, which is great. Every single day that you don't have a diagnosis or every hour that you don't have a diagnosis in these patients, actually, it's a pretty big burden on the family and the health care team.

So, we got the system in January, and we presented some of this work at AGBT. Initially when we first got it, it was shortly after holiday break. We got set up in our lab really quickly. Within a week, we were generating data, and in some of the initial training runs, we were trying out the workflow. 

We weren't really focusing too much on speed, although we wanted to try to keep it as quick as possible. And we were able to get it down to 12 hours and 10 minutes, which is great. That was great. So this is three samples at a time being run.

Here you can see just like looking at the quality scores between HE001 through HE004. Mark presented a lot of this. I'm not going to go into details, but like the overall performance is really good. Mark also showed this in kind of a different way, but just looking at performance, this is focused on SMBs, but the same thing holds true for indels. If you look at the homopolymers and high GC, low GC, it performs really well.

We took some of this same approach, and we started testing first patient cell lines. So we were looking at patients with different conditions, oftentimes that would be diagnosed in the neonatal period. So there are quite a few inborn errors of metabolism that are here.

And we were able in all the cases to detect the causative variant or variants. And so you'd see there's a list of small variant tests that we did, large copy number event cases, and then repeat expansions, also using some of the Roche developed bioinformatics solutions for repeat expansions. We were able to detect all of these causative findings.

We then took that same approach and tested some previously diagnosed patients in our laboratory, in our clinical lab. And this is just like patients that were tested and run by our standard 30X genome. And we were able to find the causative variant in all those cases.

And so all this work, like I said, was presented at AGBT and it was all generated in January. Mark kind of showed a little bit of this. So this was kind of like the starting point of you know, we have a really fast assay, but how fast can it be? And like this was, this was like in February right before the conference, we got it down to six hours and 25 minutes, which was absolutely amazing. This is just one sample. 

Okay. So, like a single sample going through the workflow, Mark showed now we could do it two hours less. So four hours and 25 minutes, which is nuts, still one sample, when we think about it,

As I said, our goal is not how do we show the fastest genome. We are interested in showing the fastest genome, so I don't mean to say that this is not of interest.

But when we think about productionization and actually running on a very routine basis, we want to run something that's going to be able to actually handle the needs of the clinic. And so for rapid genome, the most useful approach would be testing a three-sample set because you could have mom, dad, and proband. And that could be really, really informative when you're thinking about variants and inheritance. Is it a de novo variant? Was it inherited from a healthy parent?

All these things are really, really useful when determining the significance of a variant. And so when we started, after the conference, we spent some time, a lot of the work was done in the Seattle group from Roche, but then, taking this technology and actually applying it in our hands, we were able to get three samples, so three whole genomes going from DNA through variants in a really aggressive time. 

So, it was like eight hours and two minutes for three genomes to be sequenced, which is great. We even got one to be below the eight-hour time point.

We were ecstatic, fantastic. And then we kind of added some more tweaks in. So this is a constant iterative process. And now we've got it down to seven hours and eight minutes for three samples being sequenced at the same time. So this is two separate runs of trios just to show that consistently we're getting over 30x. Consistently, we're getting our time to be in this really, really amazing time frame to be able to sequence genomes.

We wanted to see, you know, it's great to have a fast genome, but can we actually detect positive variants in patients? So we tested another group of cell lines. Some of these are the same ones, but a lot of them are different.

Also, looking at small variants, a lot of these are inborn errors of metabolism and other conditions that present in the neonatal period. Not all of them are, but most of them are. We looked at large copy number events, and we also looked at repeat expansion. 

For all these, we were able to detect the causative variant again. So we're not just getting really fast sequencing and really accurate sequencing, we're actually able to find diagnostic hits in cases. So these are a set of cell lines, and this is just like looking at some of the data. 

So, the data looks really clean and really nice. So this is a single-nucleotide variant for Usher syndrome patients. This is looking at cystic fibrosis, looking at a heterozygous call. Looks really nice. This is a single. So this is a base deletion, I guess a three-base deletion of cystic fibrosis. So this is the classic Delta 508 variant, that's like super common in Northern Europe, homozygous single base pair deletion. 

So, you'd notice a lot of the propionic acidemia cases, like Mark showed it to. This is great because it's like my kind of pet disease of interest, and so I keep plugging it, and so that's why there's a lot of PCCA and PCCB samples getting tested.

This is just another example of a complex insertion/deletion. So this is in the second subunit of the propionic acidemia, propanol-CoA carboxylase gene, PCCB. In all these cases, the data is really clean, and it looks great. And these are not just like looking for events in IGV, although that's what I'm showing. These are actually calls that are being made in the VCF.

This is looking at a copy number event in the Duchenne muscular dystrophy gene DMD. This is a patient who has Becker muscular dystrophy and we're able to find the causative event here too.

There are a bunch of other examples too, I'm not going to go into many more, just to show we were able to identify repeat expansions, so this is looking at a patient that has Friedreich's ataxia. It's an autosomal recessive repeat expansion condition, which is different than a lot of the other repeat expansion conditions. We were able to see two expansions at different sizes that are in the full expansion range. So really cool to be able to see all these different use cases.

When we're testing patients, we don't really go into the VCF file, nor do we typically go searching around for things in IGV. We oftentimes will use tertiary analysis software. And so great to see that things are detected in the VCF., Great to see that we're able to see the variant, but are we actually able to identify and prioritize variants in cases?

So, looking at a sample, once again, propionic acidemia sample, this is a trio where you have both parents are heterozygous carriers, and then the proband is homozygous for this single base pair deletion. This is using the platform that we use clinically at Broad Clinical Labs, which is Fabric Genomics, Fabric Enterprise System, and we're able to see the causative variant that was prioritized all the way at the top, very high score.

And once again, this is not being used for clinical use, but if it was being used for clinical use, the next step would be basically the creation of the clinical report also able to be done. This was more like proof of principle. Can we take a sample and the VCF and load it into tertiary analysis platform? We did this kind of like under the radar, without even telling Fabric or really anybody else that we were going to do it, and it worked, which is great. 

So, it performs very similar to other short-read technologies like Mark showed. Especially when we're looking at the fast, the read lengths are comparable. They're a little bit longer in this duplex format, but overall, it's able to kind of come into different tertiary platforms.

So, obviously, we've done a lot in the last, I don't know, two or three months. And in the last five or six months, we've done quite a lot, but there's still a lot of work to do. And so Mark kind of alluded to this earlier. We have this collaboration agreement that was announced between our group at Broad Clinical Labs and the Roche sequencing team.

How do we take this from concept? How do we continue to optimize the bioinformatics for these complex genes and variant types, which is really important? We really have to streamline this. So it's fast and easy, but it still takes a person to do it right now. And so we'd like to get to automation whenever it's beneficial.

We showed one example of tertiary analysis. But as you can see at this conference, there's a gazillion different tertiary analysis vendors. And it'd be great to kind of see how this performs in lots of different tertiary analysis platforms, as well as giving vendors the opportunity to potentially tune their AI and algorithms so that it could detect the causative variants. 

Because like I said, we did this with Fabric without them even being aware that we were doing it. And then validating an end-to-end workflow. And then really the exciting part for me is taking this idea and starting to implement it for actual patient care.

So, pairing up with a partner who's already set up and really interested in this, how do we get the rapid genome to be done in one day and one shift, which is pretty awesome and kind of unheard of at this point? So this worked very little. It was actually done by me.

This is really the team that was responsible for all of the work that you saw here. The picture is the Broad team standing in front of our systems. They don't look as pretty as the ones at the booth. They're still kind of early alpha release, and obviously, this is not just work that we're doing independently. 

The Roche team has been hugely helpful, and really, it's a great partnership. It's kind of rare to interact with a group and a large corporation like Roche with a team that operates very much like a startup. 

So, I'm pretty sure Mark's whole team is sleeping in the laboratory and not doing anything except for this, but it's great. So, we're constantly pushing the envelope of what's possible. So thank you.