Welcome to the Silver Sponsor Workshop and Lunch, sponsored by 10x Genomics. Are you guys ready for some real mind-blowing science? Please welcome to the stage, Sarah Taylor.
Hello, everyone, and welcome to the 10x Genomics Silver Sponsor Workshop talk. My name is Sarah Taylor. I'm Senior Director of Applications at 10x, and I'm delighted to be here today to talk to you about what we've been working on in R&D over the past year.
I'm joined by Mike Schnall-Levin, our CTO, Patrick Marks, fellow in our computational biology team, and Ben Hindson, Co-Founder and CSO. We're incredibly excited to be here this year, and the talk's incredibly special to us because for the first time, we're going to unveil new products across all three of our platforms.
This slide contains important information regarding statements we'll be making today. Please visit our website for more details. I want to start with the why. Why are we all here today?
You all know biology is incredibly complex, but have you really thought about how this complexity limits our understanding of disease and our ability to improve human health? Unraveling the complexity of disease and separating it from healthy biology has been challenging, and I want to take a few moments to highlight some examples of this that really stand out to me. In Alzheimer's disease, there is a 98% failure rate in drug development.
This is one of the highest drug development failure rates. In cancer, checkpoint blockade, which has become a cornerstone in many treatment plans, only provides benefit to a minority of the patients treated. P eople with autoimmune diseases can go through over four years of testing before they get an accurate diagnosis. I picked these examples because these are not rare diseases. These are common diseases in well-studied, well-funded areas of research.
I'm sure everyone here will know someone or even have a loved one struggling with a neurodegenerative disease, battling cancer, or suffering from an autoimmune condition. This is real life, and this highlights that what we do not understand about human health and disease is vastly greater than what we do understand. 10x seeks to help change that.
To make progress in unraveling the complexity behind some of those stats, we need the best technologies that provide the most insights, and we need technologies that provide insights at the resolution of biology, so that's at the resolution of the cell. Not just any technologies. We need robust technologies that work reliably and reproducibly in your hands, in your labs. We've built three industry-leading platforms that allow you to look at biology, each with a unique perspective. We have Chromium Single Cell, Visium Spatial, and Xenium In Situ.
Depending on your question of interest and the prior knowledge that you have of your system, Chromium, Visium, or Xenium may suit your needs better. So Chromium is our platform for the analysis of dissociated cells and would be your choice if you wanted a multimodal readout with single-cell resolution. Visium and Xenium are our spatial platforms. So Visium is primed for spatial discovery studies with its whole transcriptome capabilities, and Xenium would be the right choice for you if you wanted to look at hundreds to thousands of genes with precise spatial subcellular resolution and the highest spatial per gene sensitivity.
Today, we're going to tell you about updates across all three of these platforms, but before we do that, I want to take a moment to talk to you about how we build products and our philosophy behind how we do product development.
There are four key areas that we've invested in and prioritized since the early days of 10x, that we believe are key differentiators that enable your research success. T hey are innovation, building things from the ground up, thinking very carefully about how we develop things in R&D.
Operational scale, being able to build these products, having the manufacturing capabilities to do that. Analysis and insights, getting you from that endpoint of your experiment all the way to your biological findings, and then support, having incredible service and support teams. First, it starts with our innovation engine. At 10x, we're hard at work developing tools that allow you to answer key questions in your research. When we start a product development journey, we ask ourselves: What are the big problems in biology?
What biological questions are you asking, and what technological advancements do you need in order to be able to address those questions? We meticulously innovate, develop, optimize, and test so that the final product that reaches your lab is something that just works. The second area we focused on is our operational scale and expertise.
A key part in the design process of any product is understanding how you're going to build it. And I'm sure everyone here can appreciate that having a great idea is one thing. Turning that into a product that you can manufacture at scale is a whole other thing. We've built our global manufacturing capabilities to ensure that the products you buy from us are reliable and robust. And there's some examples of here, this in the photos here.
You can see for Chromium, you're looking at an instrument that fabricates bespoke inserts used to mold our single-cell chips. In the middle, our in-house photolithography clean rooms that we've established to build out the Visium slides, and on the far right, the vast production lines that we've built to get the Xenium instrument to you. We've built these incredible facilities so we can scale to meet the demand that we have for our products.
Third, we've built world-class analysis capabilities to accelerate and support how you analyze, interact with, and derive biological insights from your data. On the screen, you're looking at the latest version of the Loupe Browser, Loupe 7, being used to explore some single-cell data from lung squamous cell carcinoma samples.
As you can see, 10x visualization tools allow you to easily interact with your data and derive biologically meaningful results, and we provide this capability across all three of our platforms. Finally, customer support. We have a phenomenal global service and support team who are committed to enabling your experiments, from installing your instruments and training your team, as you can see in the pictures here, to helping with experimental design and planning.
On the rare occasions where things don't go as planned, our technical and software support teams are available to help. What all of this amounts to is lots and lots of incredible research being done, exemplified with a large number of publications that have used 10x products to get to their findings.
I can confidently speak on behalf of all of 10x when I say it is incredibly motivating for us to see the products that we've designed, built, manufactured, and supported, having a real impact on your research. Over the last year, we've continued to focus on advancing our technologies to allow you to measure more biology with increased ease of use.
But something that we've heard a lot from you is that you want this, but you want this at reduced cost. Whether that's because you're a new lab starting out or because you want to scale your studies, we understand that cost can be a throttler. S ome of the things you're gonna hear from about today is what we're doing to bring the cost down. T his is something we're really focused on continuing to push on as we go forward.
With that, I'm going to hand you over to Mike, who is going to share some news we know you've been waiting for on Visium HD.
I'm Michael Schnall-Levin, and I'm the Chief Technology Officer and a founding scientist at 10x Genomics, and I'm honored and humbled to get a chance to represent the Visium development team that's back at 10x watching this right now, and take you through some of the latest advancements on the Visium platform.
If you were to design an ideal spatial discovery platform, what would it look like? Well, I think first it would be whole transcriptome, so there'd be no sacrifices to the kinds of discoveries you could make. Second, it would be compatible with a broad set of real-world samples, most importantly, human FFPE samples. Third, it would generate high-quality data, high sensitivity and specificity, but also importantly, consistent and robust spatiality. Next, it would allow you to do imaging from the same tissue section, most critically, H&E imaging.
It would have an easy workflow and logistics, and it would generate data at high spatial resolution. A s we built out the Visium platform over the last several years, we introduced a key foundational product that has enabled us to deliver on many attributes of this ideal solution, and that's the CytAssist instrument. W e realized along the way that the CytAssist was so foundational in delivering these attributes, that we made the decision a few years ago that all new capabilities on the Visium platform would come out on the CytAssist.
O ver the last several years, we've introduced multiple versions of our Visium FFPE assay, we introduced the capability to do fresh and fixed frozen tissue, and we introduced protein and RNA multiomic solutions from the same tissue section, all on the CytAssist.
At its heart, the CytAssist is built on a simple concept that upfront tissue histology workflows should be completely separated from downstream molecular barcoding. O n the CytAssist, all tissue sectioning and imaging follows completely standard tissue histology workflows. This means that it's easy to take the CytAssist and build it into existing workflows, that tissue collection can be decentralized, and that imaging and pathology annotation can be used for slide and region selection.
Then downstream from the standard histology workflows, molecular barcoding proceeds in a highly automated fashion on the CytAssist instrument. Because of this automation and the precise control that this instrument affords, this molecular barcoding can proceed in a manner that's far more accurate and precise than it could with a manual workflow.
We initially designed the CytAssist to really enable easier logistics and, and workflow, and it has done that very much so. But what we found during the few years of development at 10x on this product, and subsequently getting it out into many customers' hands, is that it is absolutely critical for generating the highest quality data consistently.
It's really critical, in particular, that during this molecular barcoding step, that you capture molecules from the tissue extremely close to where they originated. W hen you think about a spatial transcriptomic solution, the resolution of that data is not just generated by the resolution of the barcodes on the slide; it's by the entire solution that has to come together. And the CytAssist is a really critical part of that. T his is even more critical as you move to higher resolution. So we've been really excited.
The CytAssist is now out into many hundreds of customers' labs, and it's been really successful, and it's enabled science that would not have been possible otherwise, in many cases, or would have been extremely difficult. I just wanted to highlight a couple of studies.
The first is a study that's being done jointly by Joakim Lundeberg at KTH in Stockholm, and Alex Swarbrick at the Garvan Institute in Australia. This is really the first example of a multi-continental spatial transcriptomic study. Alex and Joakim are studying breast cancer, and they're studying human FFPE breast cancer samples. They've measured more than 230 of them using the CytAssist and Visium so far.
In this study, the tissues are sectioned, imaged, and annotated in Australia, then coverslipped, shipped around the world to Stockholm, where they're molecularly barcoded, sequenced, and the molecular analysis is conducted. They're using the combination of the overlay of the H&E image and the molecular data to study tertiary lymphoid structures in breast cancer.
The second study I wanted to take you through at a very high level is being done in the lab of Hanli Ji, who's an oncologist at Stanford University. Hanli's lab is using a combination of Chromium and Visium to study gastric cancer and look for gene signatures that are indicative of risk for progression of the disease. In particular, he's been applying, in his lab, Visium on a set of precious samples that are derived from people with gastric cancer or advanced gastric cancer precursors.
These sections were only available already pre-sectioned onto slides, but with the CytAssist, they were analyzable, and they generated high-quality data. W ith Visium V2 and the CytAssist, our customers have been able to generate extremely high-quality data from a broad array of samples, and data like this being shown here with this human FFPE colorectal cancer sample. But we've consistently heard one thing from all of you: You want all these benefits, but you want that at the highest resolution possible. N ow we're really excited to be able to put that capability into your hands.
Here, you can start to see a glimpse of the power of HD. HD delivers all the benefits that I've been taking you through, but does so at a massive increase in resolution, going from 5,000 features in our on-market Visium product to 11 million features in HD.
I'd like to formally introduce Visium HD. Visium HD generates whole transcriptome data from human and mouse FFPE samples. Features are printed in a continuous lawn at 2-micron resolution with no gaps in between them. In a lot of the data I'll show throughout these slides, by default, we're binning those to 8 microns.
HD is built for FFPE, inherits the advantages of the CytAssist I've been taking you through and generates high-quality data, and it comes with updated versions of our Space Ranger analysis software and Loupe visualization browser.
I'm gonna take you through the top benefits of the HD product, but I also encourage all of you who haven't to go see poster 654 by a true Visium HD badass, who's spent the last few years of her life making sure that this product would come out at high quality, Monica from 10x. T he first value proposition is that HD is built for H&E, for FFPE.
When we designed the product, when we optimized the product, and as we've tested and validated the product, we've done so on a broad set of human and mouse FFPE tissues and blocks. And importantly, the protocol that comes with HD is nearly universal, meaning almost any tissue is going to work with that with no optimization in customers' hands.
If you'll indulge me for a second, we thought some of these images were so striking, we wanted to just step through them because they're really incredible to look at. T his is human colon and human colon cancer. This is human lung cancer and human prostate cancer, and again, this is the unsupervised clustering that's being shown in all these cases. This is human tonsil and mouse embryo.
When I look at these for the first time, I think they almost look like alien worlds or something, especially this human tonsil. It's kind of amazing to think about just being able to walk through that and just make all the discoveries that you could.
But down here on Earth, real-world samples have serious challenges, and we've spent a lot of time on Visium HD, trying to make sure that it's going to work in the real world on the kinds of samples that you're going to encounter. H ere's an example where we've taken an FFPE block, and we've either freshly sectioned that or we've stored that archivally on slides for multiple months via several different methods, and you're getting consistent, high-quality data and nearly matched performance across all these methods.
The second value proposition of HD is that it's a true high-resolution solution, and as I talked about earlier, that is a lot more than just printing barcodes at a high resolution on a slide. It's a complete solution.
If you think about the first generation of spatial transcriptomic assays, which the first generation of Visium was, and all non-Visium spatial transcriptomic approaches are currently limited to, there's basically a manual release of the molecules and capturing of those on the surface of the slide. W hat that means is that these molecules will actually typically diffuse or flow some distance away in a somewhat uncontrolled manner. W hen there are gaps in between the barcoded features, that can actually be exacerbated.
But in HD, built out on the CytAssist, this is an extremely controlled process, and that means consistently, you can capture molecules extremely close to where they originated in the tissue... So we've been really obsessed with this aspect of the data quality during HD development. W e used it to actually inform not just testing, but key developmental decisions.
As we made architectural trade-offs and key detailed trade-offs in how we built the assay, we were, we were paying attention to the spatiality, and I'm just gonna show you one example of this here.
This is mouse intestine, where we're looking at the crypts, and inside those crypts are Paneth cells, and we've used the ability to overlay the H&E image and the molecular data to have someone manually annotate regions that have those Paneth cells and then look at two genes, Lyz1 and Defa21, that are only expected to be present in those, in those Paneth cells, and you can see that confirmed here. That's just one example.
We've now done that across a broad set of human and mouse tissues where there are well-known marker genes expressed in either given morphological regions or given cell types, and where you can take the H&E and the molecular data to confirm that the relevant marker genes are where you would expect them. I n these cases, I'm just showing you one or two genes in each of these samples. R emember, these are being drawn from data that actually contains nearly 20,000 genes in each of these experiments.
The third attribute of HD I wanted to take you through is the sensitivity. So this is a plot that we showed at last year's AGBT, and this was a comparison of Visium V2 on the CytAssist to our very first Visium V1, where we're looking at the sequencing saturation curves.
You can see the huge improvements we've made in multiple tissues over the years in driving that sensitivity higher. Here's now a similar comparison on that sequencing saturation curves for Visium HD versus Visium V2. You can see that we were able to incorporate all those advancements we made and basically get matched or even slightly improved sensitivity with this massive increase in resolution.
The fourth attribute I wanted to highlight is the ability to generate H&E images from the same tissue sections that we generate molecular data from. You've seen that I've already used this a lot throughout this talk, and I really want to emphasize it's the exact same tissue section. To highlight why that's so critical, we showed this example here on the slide.
If you look at the left, you can see in the H&E, morphologically annotated as an individual plasma cell that's present in this tissue section. I n the overlaid molecular data, you see this J chain marker that's present right where that, that plasma cell is. I f you go just one adjacent tissue section over, that cell is missing and the molecular data is missing.
The ability to do H&E imaging and molecular data on the same exact tissue section is really, really useful to be able to validate your molecular data, but it's also extremely intriguing when you think about the future of digital pathology. I'd say perhaps the most important attribute of Visium HD. First, that HD is whole transcriptome. So nearly 20,000 genes are being measured across all 11 million features in every experiment.
I have just a quick video that one of our scientists created as they walked through Loupe, our browser on the standard output from Space Ranger. H opefully that will play. There we go. Okay. T his is a mouse embryo, and you're seeing the H&E image and then the molecular data being overlaid on top. F irst, you're seeing the clusters from the unsupervised clustering, and each one of those clusters is being driven from the underlying whole transcriptome data.
N ow you're seeing a family of keratin genes first combined and then individually separated. This is a large family of genes, and you can see the different spatial patterns of each of those genes. And when you have whole transcriptome data, there's no need to limit yourself. You can look across this broad set of genes and across all gene families.
Now I just want to quickly take you through a glimpse of some of the biology that you'll be able to do with HD, and this is drawn from some work that our applications team is currently doing on HD and planning to put out alongside data from Xenium and Chromium in an upcoming preprint.
H ere they're looking at a human colon cancer FFPE sample, and on the left is unsupervised clustering from the HD data, and on the right is a tumor deconvolution using the single-cell RNA-seq, and then that green color is the tumor score. T he first thing you can see is that there's a really ask confirmation of where the tumor is and that clustering that you're able to find from the unsupervised clustering.
That's not just true at the gross morphological level, but in the details of the features, if you actually look closely. A gain, just showing you a quick glimpse, what the team has been doing is a tumor periphery analysis. T hey're taking those regions of tumors. They've zoomed in here in a 250-micron region within that, and then they're looking at different gene markers and how they vary as you move different distances from the tumors.
You can see here two examples: one, IGKC, which is a marker for B cells, and you see that's actually largely excluded from the tumor and really only showing up once you move some distance away. And then CEACAM5, which is a marker that's present within the tumor. So hopefully, this short talk has gotten you guys as excited about HD as we are.
As usual, from 10x, this is just the start, and we have a lot more planned for this platform. We would love to hear from all of you on what you most want as we build out capabilities on this platform. A s it says here, we're taking pre-orders, and we expect to be shipping HD this quarter, very soon, and we really can't wait to see what you can do with this new capability. With that, I'll hand it over to Pat.
Hi, everyone. I'm Patrick Marks from the Computational Biology group at 10x, and I'm really pleased to represent the Xenium team, give you an update on what we've been up to and our roadmap for 2024. W e launched Xenium just over a year ago, and it's been awesome to see the enthusiasm that our customers have had, and it's great to see the tweets of the instrument being uncrated and installed.
But the critical question is: Does it run robustly, and does it generate good data? So we're most excited to see the snippets of beautiful images coming out from these labs shortly after the instrument got installed. That level of performance for a brand new instrument in a new category is a huge testament to the team, the talent, the dedication of the team that built this at 10x.
That performance has also led to a rapid adoption of Xenium, and we're proud to say we sold 250 instruments last year. Many Xenium users have been able to get through the cycle of install, data generation analysis, and through to publication within the year, again, speaking to the readiness of the platform to contribute to scientific progress. Late last year, we saw the first benchmarking studies that started to compare the performance of different in situ platforms, and importantly, these were studies done with commercial systems in the user's own labs, and they were done on FFPE.
FFPE is the standard sample type in pathology, which makes it the most interesting sample type, but the RNA quality in FFPE samples can vary widely, so you really need to do matched samples when comparing platforms.
The first study was from Sami Farhi group at the Broad. They used a really interesting TMA that had 170 cores covering 36 patients and six cancer types. That yielded a really large N of distinct real-world FFPE specimens. Xenium emerged pretty clearly as the winner in terms of the sensitivity and specificity of the assay in that comparison.
The lead author of that study, Huan Wang, is presenting tonight at the technology session, so I encourage everybody to check that out. The second study is from David Cook, Lou Martelotto, and Jasmine Plummer. In their study, they've generated an orthogonal single nucleus reference using the Flex protocol and compared different platforms to that reference data set.
You can see Xenium got very good concordance with that reference across the dynamic range of expression, and that leads to the high-quality marker gene analysis you see on the right. I'd encourage you to check out these publications. There's lots of insights into how these platforms compare.
Throughout last year, we introduced a big family of pre-designed targeted panels covering a wide range of tissues and application types. We also really streamlined the customization workflow. It's critical to add content to these panels to get at the relevant biology, and so we've made that a very fast and easy process. We've also opened up more advanced customization modalities. You can make a fully custom panel on new species.
We'll help you with that, and we'll also help you design targeting targets to more interesting, exotic targets, such as isoforms, viral, bacterial, and, and other kinds of targets. W e've had some great luck, customers have had some great luck on some of these interesting areas.
Okay, now I'm really pleased to get into the roadmap. We have 4 exciting launches planned in 2024 that will expand the capabilities of Xenium. First up is a multimodal cell segmentation capability. It's gonna be shipping in Q1. Second is our 5,000-plex panels that are gonna be available in the middle of the year. Third is an integrated RNA plus protein inline, multiplex protein capability, and finally, a family of very flexible 2,000-plex panels in the back half of the year. So let's dig into it.
We did a quite a big development effort to generate our cell segmentation solution. We screened a very large number of clones to get the markers we needed, and we also combined that with a lot of optimization across many different, you know, sample prep and staining conditions to get to a high-quality stain that retains the RNA data quality.
Once we had that stain, we went and manually labeled 5,000 image patches across 10 tissues, 150,000 cells, to train the machine learning to interpret that signal. So I'll walk you through what that signal looks like. So here's the DAPI image that you already get.
We actually put quite a lot of work into the DAPI segmentation algorithm, and it gives great boundaries of the nucleus, but then currently, the platform will do an isotropic expansion out from those boundaries, and that's sort of a rough guess of where the cell lives. So with the multimodal cell segmentation, the primary data source is this boundary stain, it's targeting broadly expressed cell surface markers with an extra focus on leukocytes and epithelial cells.
The algorithm that interprets this signal does not need a nucleus, it's able to capture multinucleated cells or cells that are lacking a nucleus. C ritically, if there's not good signal in the boundary stain, it will decline to call a cell rather than hallucinate sort of an arbitrary boundary. If there's not a good boundary signal, we'll fall back to an interior stain.
We've got two channels, one targeting 18S ribosomal RNAs and one targeting intracellular structural proteins. That's used to constrain the expansion of the cell boundary away from the nucleus. F inally, we'll fall back to just a fixed expansion from the nucleus if there's absolutely no signal.
Here's what the results look like. You can see the elongated cells. You can see that the morphology, it doesn't extend into the interstitial space. And if I overlay the old boundaries in blue, you can see that we don't cover that interstitial space, and it matches to the transcript signal much more accurately.... We've tested this across a really broad range of human and mouse tissues, FF and FFPE, and we're seeing great compatibility across all the samples that people are using on Xenium... Here's an example of why this is critical.
I'm highlighting some fibroblasts. When with the incorrect boundary, they were getting some signal from neighboring pancreatic islet cells and were being misannotated. When you get the cell segmentation right, you get the elongated morphology, and you get the right cell type assignment.
Xiao Yan from the tech team is gonna be at poster 537 and can take you through this and some other interesting analysis of these results. T he cell segmentation will say, will ship as an add-on kit that's compatible with the existing panels by the end of Q1. It's as you saw, it's validated in a wide range of human and mouse samples. The software will run onboard the instrument as part of the primary analysis, and future Xenium assays will be enabled with this capability, starting with our upcoming 5K plex panel.
Let's switch to that. You have to... We thought really carefully about the scaling issues that in situ technologies face as you increase the plex. You want to maintain high sensitivity and specificity, but more plex means you're tagging more transcripts per unit area. W e've completely overhauled the molecular constructs, the optical barcoding designs, and the algorithmic stack to work reliably at this much higher total transcript density. W e've done this while maintaining good throughput. You'll be able to run 2 fully loaded slides, totaling 4 square centimeters in 6 days.
Here's a sampling of the data that you can expect with the 5K plex kit. On RNA-rich, fresh frozen tissues, you can expect to be in the 1,000 transcripts per cell range, and with challenging FFPE tissues, you'll be in the many hundreds of counts per cell.
These all use a uniform sample prep, so you can even package multiple different tissues on the same slide. Here's data from a really challenging DV200 of 39 sample of a lung adenocarcinoma, and we are tagging two broad tumor epithelial cell types and a host of immune cell types. If we zoom in, we can see some really interesting examples of cell interactions at a very small length scale.
On the top, you've got some T regs interacting with fibroblasts. On the bottom, you've got a macrophage interacting with a couple of CD8 T cells. So this sensitivity and resolution really gets you a very intimate picture of a tumor being interacted with by the immune system.
There's an interesting emerging application for screening a large number of cells and getting some morphology data at the same time, and Xenium is an interesting tool for this kind of use case. H ere we deposited PBMCs onto the Xenium slide and profiled them with the 5K plex kit.
You can see some interesting variation in the morphology image and the dense RNA data, and that yielded the classic PBMC UMAP plot you see on the right. Yutaka Suzuki at University of Tokyo has done this on his Xenium and was able to get 600,000 cells on one slide, which translates into a very low cost per cell. W e're excited to see how that use case develops. And here's what happens if you fill up the slide with a fresh frozen sample.
This is a liver sample that's 2.3 centimeters squared. We got 1 million cells, median transcripts per cell 1,500, for a total of 1.7 billion transcripts. We have a poster with some more details of the panel, this data set, and some interesting biology we found, please check out that poster and get ready for a big step up in the density of your Xenium data. W e're really excited to deliver this 5K plex panel in the middle of this year. In the second half of the year, we'll be introducing an integrated protein and RNA profiling solution.
You'll be able to choose from a growing catalog of antibodies that have been titrated and optimized for use with Xenium, and the data will be seamlessly integrated with the cell segmentation and RNA results and compatible with all the downstream tools. H ere's a zoom-in of some germinal centers in the lymph node, and I'm showing you four of the markers from the panel that are tagging all the major cell types in this, in this germinal center.
Here's another four markers. If you focus on those CD68 macrophages in yellow, you can see they're sort of forming these punctate structures. Now, if I flip over to the RNA clustering results that are derived from the same tissue, you can see the great concordance of the cell typing and the alignment of this data.
The combined assay maintains H&E compatibility, so you can import that layer as well to get a very rich view of your tissue. O f course, we've tested this panel on a broad range of tissues. We expect the protein capability to be applicable to the same wide range of samples that you're used to with Xenium. W e're really excited to deliver this capability in the second half of the year.
I don't have time to do it justice, but we make a huge investment in the software stack for Xenium, and I wanted to highlight two key aspects of what that gives you. Number one, all the primary analysis of Xenium happens on board the instruments, fully automated.
When the run finishes, you have results that you can take into Xenium Explorer or into downstream, analysis tools like Seurat and Scanpy with no additional, analysis steps.
S econd, the Xenium Explorer visualization tool that you're seeing on the right is really built from the ground up to explore the combination of the transcripts, the cells, and the morphology. T o zoom smoothly from, you know, the micron scale, where you're looking at individual transcripts, all the way out to the centimeter scale of a whole tissue. And we've got a exciting roadmap of additional features that'll kind of round out the feature set of Xenium Explorer coming this year.
To recap, we've got our multimodal cell segmentation solution coming in Q1, 5,000 plex panels coming in the middle of the year, multiplex proteins capability coming in the second half of the year, along with a more flexible family of 2,000 plex panels that will have slightly higher throughput. L ong term, we know that there's, like, a huge family of samples and studies that would be very interesting to do on Xenium, and we're really committed to driving up the scalability and driving down the per sample cost of the system, so that more of those studies can come onto Xenium in an affordable way.
10x is really excited to deliver on that long-term vision. So thanks very much for your attention, and I'll hand it over to Ben to cover the Chromium platform.
Hi, it's an honor to be here on behalf of the Chromium development team. L ast time I was here on stage was 2015, when I had the opportunity to introduce our first product, I t's a pleasure to be back here. And since then, we've been on this incredible journey of innovation for single-cell analysis.
We've invested a lot of money across all the different applications, but most importantly, is the impact that we're having on your research, and pleasure to be on that journey with you. But we like to push the limits of what's possible, and so we're not stopping by any means. And so today I'm really pleased to share with you some of the latest updates and developments on our Chromium platform today.
Just as a quick refresher, there's two categories of chemistry that we use for our single-cell assays. One's reverse transcription-based, the other is probe-based. There's different strengths for why you'll want to choose each platform. If you want the broader set of information like isoforms, SNPs, species-agnostic chemistries, or low cell input requirements, the RT-based chemistry is a good go-to. For our probe-based chemistries, called Flex, offers easy sample batching, the lowest cost at scale, and it's the only way to run single-cell FFPE tissues. Interestingly, too, it performs really well with low-quality input samples. So let's spend a few minutes talking about Flex.
One of the key attributes of Flex is its fixation compatibility. This enables you to batch samples, ship them, reduce logistical burdens, enable plate-based screens, unlock new biology, which makes your life easier.
It's real, and it works, and so rather than staying up Friday night to process those late-night samples, you can fix them, come back at your schedule the following week, collect them, analyze them in a batch, and get top-quality data. We introduced Flex for FFPE at the last AGBT meeting. We're a little conservative because FFPE is a big, bad universe of sample quality out there.
But a year later, we're really happy to see the results that are coming out of people who have tried Flex on FFPE and getting really high-quality single-cell data from archival tissues. So you've got precious sample sets, and you want to get single-cell data out of them from archival FFPE, Flex is the answer for you. F inally, Flex offers massive scale.
In a single run of our Chromium chip, you can analyze up to 1 million cells at a time at the lowest cost possible. T he lowest cost per cell at massive scale to enable new applications. This is enabled by our built-in bar coding and by overloading our chips to get this massive scale at low cost. So if you're thinking about large biomarker screens, atlasing, CRISPR screens, Flex is your go-to product.
In 2024, as you think about what to do this year, there's new things that are available too. Working together with our partners, Proteintech and BioLegend, intracellular protein is now a possibility on Flex, and even phosphoproteins are detectable with this technology. A demonstrator protocol will be coming out soon on this capability.
Customers can customize the Flex assays by adding probe sets to the ones that we already provide for a gene expression analysis. Carolyn has a poster 626, showing the detection of 7,000 guides detected by Flex while simultaneously interrogating the whole transcript and gene expression, and it's also compatible with protein-based analysis.
This one's super cool because whole blood fixation, this was the nice from customers, where you can now take a tube of whole blood, fix the whole blood in that tube, and now take that back to your lab and at your convenience, deplete the red blood cells, analyze that with our Flex assay and get as good a data as if you had isolated those cells at the beginning of the collection. So a really nice capability expansion on the sample side.
It's an industry first, and demonstrated protocol is up and live right now, so check that out and give that a try. Fixation is not just a thing for our Flex-based assays. We've also heard the need for being able to fix our RT-based assays, and so that's something that we're enabling this year. So upstream of our RT-based products, whether that's our 3 prime assays or our 5 prime assays, you can now fix those samples, and here's data showing that after day 0 versus day 4, you're still getting the high gene sensitivity and even productive VDJ spanning pairs using a demonstrated protocol that we'll be releasing very soon.
So fixation compatibility with RT products is now a possibility. We'll get that to you very soon here. So I mentioned we've been hard at work on Chromium.
It's not just Visium and Xenium. We've also been innovating in the background intentionally on the Chromium assays. W hen thinking about Chromium, we've kind of went back to the first principles, and we've reengineered some of the components from the ground up.
Here's a video showing just some of the internal capabilities of us, manufacturing the very first step of our microfluidic chips that are the heart of our Chromium assay. These are done in with ultra-precision machines. They're done with submicron tolerances, and this is something that's done inside of 10x to make thousands, tens of thousands of chip, chips with a new architecture that's gonna work every time in your hands. So not just the chips, we've also, also gone back and revised the chemistries from the ground up.
I'm here to introduce a new massive upgrade on our Chromium platform, and we're gonna call that GEM-X. Today is an exciting moment as we launch GEM-X. It's the next era of our single-cell solutions. It incorporates all the latest technological advancements, it's engineered for the best performance, and it seamlessly integrates with your Chromium X or iX instrumentation.
Sounds good. Well, what can it do? Well, we're going to release two new versions of our 3' and 5' assays based on GEM-X technology. H ere's the performance characteristics you're gonna be able to expect. Much improved sensitivity, up to 2x more genes detected compared to on-market versions. It's built to scale, so twofold increase in cell throughput per lane, a twofold reduction in the cost per cell, and also enhancing the data quality by reducing the multiplet rate.
We already have higher cell recovery with our existing technologies, and this refers to the amount of cells that you load versus the amount that you recover at the end of our process. It's getting up to even highest levels, almost theoretical, where you can now recover up to 80% of the cells that are loaded into the system.
M ost exciting too for me is the improved robustness. This is one of the most important things where you can have a high-stakes experiment, you can go to our instrument, you can process your precious samples with the redesigned microfluidics. I think you'll be really delighted to experience the improved robustness of this platform. This is all powered by our GEM-X technology that we've been innovating on for the last number of years here. Let's take a look at some data. Much improved gene expression sensitivity.
This is looking at adult mouse brain nuclei. Up to 80% improvement from these new assay kits. Again, a boost in the detection of the rare transcripts. So looking here at human PBMCs, you can see the sensitivity is boosted here significantly for our GEM-X product over the existing on-market products. Just to show you some data collected on 120 runs in our own lab to show you that the increased cell recovery is up to 80% across most of the assays on our new GEM-X technology.
Of course, with more cells and higher quality at a lower cost, you can choose to run 20,000 cells per lane, recommended loading input. You can get 2x lower multiplet rate and a twofold reduction in cost.
Now, neutrophils, delicate cells, may be hard to detect, but that's a thing of the past with GEM-X. We've improved the technology even further to give you even more confidence to detect neutrophils at the highest quality and the highest quantity. Shown on the left here is our on-market product, and the UMAP on the right is the three prime assay, showing that detection of neutrophils with our GEM-X products.
So let's keep going. If we take a look at four key marker genes that the HCA defined as the neutrophil markers. They're really robustly detected with our GEM-X platform here. R eally exciting development to be able to confidently detect the biology in your samples for, for, you know, precious cells like neutrophils.
When we take a look at a renal cell carcinoma sample, on the left comparing our V2 assay with our GEM-X assay, we're detecting more cells 'cause we're loading more and we're detecting more. A lso, when we take a look at the transcription factors that we're detecting here, shown on the right, we can see a much better sensitivity and detection of some key transcription factors that are important for renal cell carcinoma.
It's applicable for getting more information on VDJ information out of our assays. Here's that same sample. We're detecting similar T cell populations. GEM-X detects more VDJ transcripts and reveals more unique clonotypes with productive VDJ pairs. R eally exciting developments here for more detection and more confidence, more detection of the underlying biology.
Happy to say that wrapping this all together, the new 3' and 5' GEM-X kits are open for pre-order today. We plan on shipping these in Q1 of this year, and it's all compatible on the current iX or X Series instrumentation. GEM-X is the future of our Chromium single-cell assay platform. We have a healthy roadmap, and we've got more innovations to come. We'd like to keep you posted on those as those develop.
We like to think about how we can scale and continue to democratize single-cell analysis. We're offering the highest performance. I've shown you how it's built to scale. The application breadth is really unmatched. It's compatible with a whole range of different sample types, including FFPE fixed cells, and also, it's also compatible with all these different multi-omic modes of analysis: proteins, cell surface markers, chromatin, and gene expression.
It's our promise to be able to get single-cell analysis into every lab across the world for those who haven't yet experienced the power of single-cell analysis. And to do that, we're going to drive the cost of single-cell analysis down over time to about $100 per sample.
I should mention, too, that those kits that are available today for pre-order, they're also gonna be delivered at a lower cost than our existing on-market products. I hope you're excited about the developments that we've shown here as much as we are, and I'm happy to. I'll be so happy to see how it works in your lab, and, thanks for your attention, and good luck with your single-cell analysis. Thank you. With that, I'll hand it over to Sarah. Thanks.
Thanks. Okay, thank you, Ben. W e've been really excited to share all these updates with you today. As I said at the beginning, we talked about more biology and increased ease of use, and you've seen more biology demonstrated by higher resolution, more cells, and increased multiomics on some of the platforms. We're working on the increased ease of use.
We make sure all our workflows are very straightforward and easy for you to execute in your own labs. And you just heard about from Ben on some of the things we're doing to bring down the cost to make it more accessible to you all. B efore we finish up, just wanna point you to the posters that we have here today. Lots and lots of different things are from our scientists in R&D.
All of the people presenting these posters have worked on these products, so they will be able to give you very thorough answers to your questions. If you would like one of these absolutely fantastic T-shirts, you are very welcome to come by our booth after the talk and pick one up. I just wanna close on career opportunities.
Lots of open positions at 10x, and so if you are interested in joining me and the rest of the team, then please do get in touch with our recruiting team, who can talk to you about what we've got available. Thank you very much for your time. Hope you enjoy the rest of the conference, and go and see the bronze sponsor talks.