All right, good morning, everybody. Appreciate everyone being here in person, and for those online, welcome as well. I want to kick off today, give you a quick overview of what we're going to accomplish during this presentation, and then I'm going to try to set the stage with sort of the complexity of the proteome and the challenge we're trying to solve that I think will then tie nicely to what you're going to hear today, both on the hardware roadmap side, but also what we're doing on the chemistry side. Again, I'll kick off. Todd will take you through Proteus, where we're at with that program, the data underlying that technology, and where we're at, and why we feel very confident in sort of the leap it's going to make in performance.
Also give you a small peek into what we have going on on the technology roadmap. Proteus is actually the preponderance of the spend today. It's the majority of what we invest in R&D, but we do have modest investments into that longer-term view, and Todd will help you see what those are and how they intersect. John will get into how we're going to get to 20 amino acids and how soon will we get there, and he'll give you details on that. Brian will sort of cap it off with, and how do we translate all of this in this toolkit with this core technology for kinetic detection into post-translational modifications. I'll close out with, how do we sort of get to launch? What milestones should you expect to hear from us over the course of the next year?
To kick it off, high level, proteomics is sort of a massive market. There's a lot of different segments in the market. Today, with our first-generation technology, we play in academic research. We have some pharma biotech, defense, and even more recently in agriculture. It's a huge market, a lot of different applications of protein detection technologies, but it's also extraordinarily complex. I think we often talk about the proteome or proteoforms. At the basic level, you have about 20,000 proteins, but when you take into account all of the post-translational modifications, the isoforms, the single amino acid variants, you get into the range of sort of millions of proteoforms that you have to tackle. It's an immensely challenging problem, and it's not specific to a single disease.
I think, obviously, in the market today, there's a lot of noise around tau protein and Alzheimer's disease, but proteoforms play a role in many disease areas. They play a role in cardiovascular, they play a role in cancer, immunology, and just many fields. I think the key is this means affinity-based approaches are going to be nearly impossible to scale to this type of complexity. Think about 20,000 proteins and some of those proteins having multiple different modifications. The number of reagents you would need is just sort of a daunting, almost impractical task.
While developing a single molecule protein sequencing technology is not easy, when you achieve the level of capability you should be able to do with this, and I think the R&D leaders will demonstrate today, it is applicable to this challenge at this scale, the scale of all of these different proteins and all of these different changes. Here's a good example. I think tau protein probably has the most visibility in our market today. Obviously, a phosphorylated version of this protein is a diagnostic biomarker in Alzheimer's. It's a target of a therapeutic. It gets a lot of attention. If you go into the literature and you look, and this study here is from Dr. Kelleher at Northwestern University in Chicago, what they've shown is it's way more complex than just that. There's probably many more biomarkers to mine inside of this protein.
You see here some of the readout they have, everything from phosphorylation to acetylation to methylation. Multi-PTM across many proteins is the scale of the problem that's trying to be solved. Variants also matter. Here's a good example in the clinical side. In this case, sickle cell disease, but there's a range of hemoglobinopathies, which are all single amino acid variants that lead to a different phenotype. This is a great example of sometimes it's a post-translational modification, sometimes it's a single amino acid variant. If you think about sequencing as the method, reading out the amino acids, reading out the PTMs, it gives you that breadth you need to really cover the range of complexity of not only biology research, but diagnostics. What happens today with this complexity?
What happens is essentially labs own a lot of specialized equipment, and most of those pieces of equipment cost in the range of sort of a million dollars. It ends up being a very centralized, core lab-heavy model. Those instruments, while very powerful, many of them have pretty laborious manual laboratory workflows, the data analysis, a lot of custom pipelines and different things. Again, it sort of centralizes the market to large, well-funded, sort of high-infrastructure core labs. When we think about Proteus, that's really what we think about. First is the core technology. What's the best way to tackle this daunting problem of 20,000 proteins and millions of proteoforms? Again, I think the R&D leaders today will do an effective job of helping you understand why we think single molecule protein sequencing is the way to do that and do it in a protein-agnostic way.
Automation is a big part. The data analysis, the laboratory workflow. Today we'll give you a little bit of a feel for how much automation are we bringing to the table here with the Proteus platform as it compares to our current Platinum Pro. Affordability. These machines, as I said, run up to $1 million or more. We think there's a way to do this and not be up in those price points. We think there's an opportunity to make this more accessible to a broader number of laboratories, both here in the United States and globally, and really bring deep protein analysis to the masses. Today, where will we focus? We'll start with Todd. He'll take us through the Proteus platform.
He'll give you a little reminder of what's the leap we've made in architecture, why did we make that leap, and get you into some of the actual data that we've been generating in our development program, and again, give you that view for, and then how does that intersect with our long-term roadmap. John will take us through the path to 20 amino acids. I think we frequently get asked, when are you going to achieve that? What's it take? Is it possible? We've had these questions over the last year or so, and I think John will do a really effective job taking you deep into how do we do this, why do we think we've hit an inflection point, and what's that mean in terms of our timeframe to deliver.
Brian, on the PTM side, what are all the capabilities and all the data that sits inside of this sequencing technology, and how do you apply that to get really broad PTM profiling? I'll round out on Proteus and address a couple of the questions we've heard. What did we learn from commercializing our first-generation technology? How does that factor into technology choices we made, specs, different things we're aiming for with Proteus? Really, how do we see ourselves going to market and give you a view of sort of both technical and commercial milestones you should expect from us over the course of now till the launch at the end of 2026? With that, without further ado, we'll bring up Todd, who will take you into the Proteus program and the long-term roadmap.
Thank you, Jeff. It's really exciting to have this presentation today. It's been a lot of work that the technical team has done over the last year. We've been executing on the vision of the Proteus program over the last year, and they've made a ton of progress, and it's really a privilege to get to come here and share it all with you. Before we talk about the status of the program, I wanted to spend some time talking about what Proteus is and just remind everyone why we're going down this architecture change. The Platinum instrument and system is based around an integrated consumable device that has both a fluorescence lifetime imager and the wells that we immobilize the peptides into. There's about two million wells in that device. What we're doing with the Proteus architecture is we're repartitioning the system.
We're taking the complexity of the imaging system out of the consumable and putting it into the instrument. That makes the consumable much simpler and much more scalable. We wanted to get on that path of really scaling the technology much more than we could do with the integrated device. The instrument then contains that imaging system that supports that simple, scalable consumable, and at the same time, we're taking the opportunity to add some workflow automation to make it easier for users. That new consumable we call the Kinetic Array, it's based around a low-cost fused silica dye at its core. It's a passive device. There's no active components on it. There are 80 million wells per device in the first implementation, and that compares to the two million in Platinum, so it's a big scale-up in the number of wells that we can support.
The architecture, though, supports scaling to billions of wells, and that's one of the reasons why we made this change. That simple glass device then gets encapsulated in a plastic assembly that provides four flow cells, so each one of those flow cells contains 20 million wells that can be interrogated, and there are features on it to support automation in the system. The instrument then contains this high-performance imaging system, and that's really what enables us to use this very simple passive consumable. In addition, the instrument contains this liquid handling system that allows that workflow automation, and it provides the capability to do more advanced workflows that could provide deeper insight that we wouldn't want to burden the user with. Where are we at?
At this point, over the last year, we've really focused on retiring technical risk in the program and figuring out all the specifications for the product. We've done that through a series of prototypes that we've used to mature all the critical system components and retire those risks. The integrated design is complete at this point based on all the learning from those prototypes, and we're building the first systems now. Through those prototypes, we've really been able to develop all of the new elements of the system and make sure they're working and get them working well. For the liquid handling system, for instance, we have a prototype liquid handling system. It's the exact same architecture as what's in the final product, and in fact, it shares a lot of the same components with the final design.
With that, we've automated our full sequencing workflow and demonstrated that working. We've implemented all the new elements that are required to make that work. What I'm showing here is data from experiments that we've run with the automated workflow that show that it's much more reproducible than the manual workflow, which is a lot of what we would expect from an automated system. We've been able to make things much more reproducible. With that kinetic array, we've done a lot of development there. We just had some simple prototypes last year at this time. Now we have a wafer fabrication process established. We've optimized the well structure and design for the best performance in our system. We've designed this whole thing from the ground up to be scalable and low cost. Those wafers have to go through some functionalization after they're built.
All of that is done at the wafer scale and processes that can be done in bulk and scaled to very high volumes for low cost. The dye size that we're using right now is the same size and geometry that's going to be in the final product. So we're already operating at the scale that we need in the final product. We've already proven all of that out. And then we have that packaging process with all the features we need for automation already developed and tested and working together. The imaging system was a big area for us. This is where we're taking all that responsibility of producing the data from the peptides and collecting it, and we're putting all that into the system. It's a big change to what we do architecturally, and it required a lot of development.
We've built multiple versions of prototypes of the imaging system. They're all the same architecture as the final product design. They have all the same features. They do the detection the same way. The only difference is they just use commercial off-the-shelf components instead of some of the custom optics that we have in the final product that gives us the full field of view that we want. With that imaging system and with the dyes that we've developed that work with it, we've been able to demonstrate, really validate the transition from lifetime detection that we do on the Platinum product to detecting in color in Proteus. This was a big transition for us. It's completely proven out. What I show on the right-hand side are you really see eight separable clusters in a two-dimensional space of color and intensity.
Those represent eight different dyes that we can have in the system at one time, and we can distinguish them from one another. That means we can detect the molecules they're bound to accurately. Just for reference, in the Platinum system, our latest kit needs six, and here we're demonstrating eight, and there's actually plenty of room there for more. This is a very accurate and extensible space that supports pretty much whatever we want to do with our assay going forward. We're very excited about that proof point. Not only have we used all of this work to develop these different components of the system individually, but we've brought it all together. For the first time at the end of Q3, we brought all these elements together and performed our full dynamic sequencing assay. This is just an example trace right here.
This is data very much like we would get from a Platinum system, all with all the new elements. The new imaging system, the new consumable, the new dyes, and the software that does the reconstruction, all working like it will in the final product. A huge de-risking of all the new elements. There are not really a lot of big technical questions out there about whether or not we can do this. It is really just a matter of execution from this point. Since we have been sequencing over the last couple of months, we have not really had any time to do optimization yet of the system.
We're really still using the same reagents and software for the most part that Platinum uses, but we have been able to measure the performance relative to Platinum so that we have a baseline for where we're going to be not just from making it work, but a performance standpoint. I'm very happy to say the performance is meeting our expectations of being significantly better than Platinum. For example, on the left, you'll see a plot about alignments per well. That means if I take the same number of wells in Platinum as I do in Proteus and then see how many alignments I get out of the end from an equivalent amount of wells from those two systems, I'm getting about 1.8 times as many alignments from Proteus as I do on Platinum.
The system is capable of producing more output from the same number of wells. On the right, those alignments, they're not equivalent. The alignments from Platinum tend to be longer and have more aligned positions, so they're more informative. You kind of get to multiply all these things together, right? We're not only getting 40 times more wells in Proteus, but those wells produce 1.8 times the number of alignments, and those alignments are more informative. It's a huge leap in the information content from the experiments that we run on the system. Again, just remind everyone, this is without any optimization. We're just running things pretty much with the same conditions that we have on Platinum. We have a lot of room to improve. Proteus consistently sequences deep into peptides.
This is an important aspect of our system that is a unique capability for long-range structural information. When we talk about PTMs and variants, it's important to have access to all the different positions, including deep into reads, and we don't see a significant limitation there that's fundamental to our system. There's a lot of opportunity to improve this, again, as we do more optimizations of the system. Here I'm showing reaching position 19 more than half the time. There's no reason we can't extend that with additional development. Very fundamentally, and what's behind a lot of these improvements, is just fundamentally better SNR for the Proteus system compared to Platinum. We have a metric in our system, which is how easy it is to detect the binding events over the background noise that we have in the system. We call pulse SNR.
I'm showing on the top right histograms of that pulse SNR for both Proteus and Platinum. Proteus is in the dark blue. That pulse SNR is about twofold higher than Platinum. This is significantly more sensitive and accurate than the Platinum detection system. This means, again, those wells produce more high-quality alignments, better accuracy, more sensitive for PTM and variants. There is another more obscure thing that may not be obvious, which is that the binding events that are hardest to detect are the short ones, the ones that are only there for a brief period of time. In Platinum, I'm just showing an example on the bottom right there where there's a tryptophan in the middle that we basically don't see it. The reason we don't see it is because the binding is weak to that particular position. We don't observe those events very easily on Platinum.
We miss them most of the time. The Proteus system, without any optimization, and in fact, the software deliberately tries to ignore these short events because they're not very informative. Despite that, there's so much signal there that the software is finding that tryptophan position half the time. There's a lot of room to improve the software to leverage the information content that's in the Proteus data that's not present in the Platinum data. This is super important because getting to 20 amino acids, getting high coverage of all the different sequence motifs, we make the binders better so that we get more coverage, but Proteus is actually giving us a more sensitive detection system. It's bringing the sensitivity up of the system so that we can get more on the low end as well.
We do not have to improve the binders as much as we would have to otherwise. A big advancement to what we are capable of seeing. Where we are is that not only do we have all these elements working together, they are sequencing, and we have proven out our strategy and the new architecture, but the data is superior to Platinum with very little optimization, and we expect a lot of additional improvements before launch. We have used the data and learning from these prototype developments to develop specifications for the final product. We are very confident in what the final product needs to be able to do to function and produce the full performance that we desire. We have already designed an integrated system. The first integrated systems are expected to be completed in the first quarter of 2026. Those will then bring everything together.
The full workflow automation, integration of the imaging system into it, product-like consumables, including for the reagents, and that full field of view supporting the 20 million wells. This will be—we're very excited to get this. I kind of can't wait. It's coming soon. In summary, compared to last year at this time, when we just had a couple of proof points and we were pretty confident about the direction, now the transition to the new architecture is completely proven. The team is just executing right now towards a product launch at the end of next year. The architecture shows significant improvement in terms of not just the quantity, but the quality of sequencing output. You're not just getting the geometric scaling of the number of wells. We're getting additional improvements beyond that because the architecture has just fundamentally better performance.
We have aligned the launch of the product to coincide with improvements in biochemistry and library prep that will drive additional performance gains. It is really like a huge leap in performance that we are going to get when this product comes out. We are very excited about that. Now I am going to transition to the future technology roadmap. When we embarked on this new architecture, it was not just to make a Proteus instrument with 80 million wells. It was to put us on a new path that was more scalable. If you look at our long-term technology roadmap, the first Proteus instrument is really at the beginning of it. It is really putting us on a path to a system that scales to billions of reads. As I mentioned before, that consumable architecture is designed to scale to billions of reads.
In fact, we could probably make the chip for the Proteus 2 instrument now. There are other things we have to do to make that work. There is about a tenfold scaling in the assay that is available just from assay improvements alone, and those are things we can do in the shorter term. There are some more advanced technology development efforts we have to do to enable that scale up to 10 billion reads. I am going to talk about some of the different things that we are working on. Just like Proteus was an advanced technology development a couple of years ago, now we have some other things that will roll out in the future once we work them out. One of those things is super Poisson loading. The system today is dependent upon Poisson loading.
What that really means is that we do not have a way today to prevent multiple peptides from loading into the wells on the chip. Because we do not have a way to do that, peptides load randomly. Sometimes you get no peptide in a well, sometimes you get one, and sometimes you get multiple. It is really only the wells with one peptide that are useful. Those are the only ones that produce useful data. It turns out that statistically, you can only get about a third of the wells, a little more than a third of the wells, populated with a single molecule in a system like this. We are working on a method right now to deliver one peptide to the majority of wells. This is about a threefold increase in the throughput of the system.
This is something that could be a kit release for the first Proteus instrument. It does not really require an instrument change, but it is something that we are working on for the future. Another thing we can do to improve the throughput of the assay is to speed it up. Since we have launched the product, the duration of the sequencing reaction has been 10 hours. Our focus has really been mostly about improving coverage and improving the information content that comes out. As we make progress on that and we have coverage, we really want to get more throughput out of the assay. One of the ways to do that is to speed it up. Instead of taking 10 hours to sequence one of those imaging spots, we can do it in five hours or two and a half hours.
That enables us to sequence multiple devices one after another and increase the throughput of the system. Without making the system a lot more complicated, it could be a drawer upgrade to the instrument. It would need a little bit more space for more consumables, but that's something we could roll in in the future pretty easily. Finally, the next big leap, because we do need another big leap to get to that 10 billion reads, is a change to the sequencing chemistry itself. The current system is dependent on this random cutting process. It's a one-pot reaction. We put in the cutters and the binders at the same time, and they're both acting at the same time. In that system, the cutting action is a random process.
It happens whenever a cutter happens to diffuse in the well and cut one of the amino acids off. I'm showing examples of that on the right. That's the same peptide detected in three different wells of one experiment. They all look different. If you look close, the order of the colors is the same, right? They're all purple, orange, blue, green, yellow, and so on. Today, that's what we use, right? We use those events in order, and the software detects the transitions between them, and it figures out which amino acids are in the peptide. Not having control of that is limiting in terms of the scale of the system. It's a little more complicated for the software. The time on average tends to be slower because we don't want to go too fast and miss things.
Also, there's no way to look away from a region. If you have a field of view that supports a certain number of wells, you can't look at one spot for a while and then go look at another spot for a while and then come back. You'll miss important information. You'll miss cutting events. You'll miss amino acid residues. In order to enable that scanning where we can take one field of view and multiply it over a much larger area, we need to gain control of this process. We have been developing a method that we call controlled cleavage. In this process, we gain control of the cutting and make it mediated by the system. In this case, the imaging process is separate from the cutting process, and it's controlled by the reagents that are delivered to the consumable.
I'm showing an example of it here. In fact, this is data that we've collected with proof of principle experiments that we're doing. Again, this is like the stuff we're working on laying the groundwork for the future. You can see that same peptide, and you can see we have these defined boundaries between the amino acids with specific cut locations. This has all these benefits. They'll be able to speed things up. We'll be able to simplify the algorithms and make them more accurate because they'll know exactly where to expect the data for any given amino acid residue. We can enable a scanning system because we can make those imaging periods. We know when they are, and we can scan during that imaging period without losing any information.
This enables something like a Proteus 2.0 instrument, which would incorporate a high-speed scanning system and microfluidic delivery to cycle the reagents to the device. This is what really lets us get to billions of reads because we can multiply that limited field of view across a large area. We also get to pack those imaging locations all together in one flow cell, which makes them much more efficient use of device area. All of this is enabled by the controlled cleavage chemistry. Something we're working on for the future, but it's really all in support of this vision of a long-term technology roadmap that gets us to billions of reads. That's what I have. I'm going to turn it back over to Jeff now. Thanks for your time.
All right. Thanks, Todd. I think to maybe summarize, first off, we came about this time last year and talked about Proteus and this huge leap in architecture. We said we would launch in two years. I think during that presentation, I had made a comment about programs like this typically run three, four, five years. There was a lot of healthy skepticism, especially from our friends in the analyst community, which we appreciate that skepticism. I think you see where that confidence came from. I think you see the methodical nature with which the team has tackled the problem. I've talked before about the depth of talent we have across every function you need in R&D to pull this off. I think that shows up when you see the diversity of the things we've been working on.
Similarly, I think we've talked about most of our investment is in product development, meaning things that will see the market be a product that customers can buy within the next year or two. We do have that small investment in the background of how we get to the end. I think we showed that data today on controlled cleavage, not because we intend to launch that product tomorrow, but because whenever you're making progress, there's always folks who are going to say, "Well, this is going to be where they get tripped up. This is where they're going to get stuck." What we want to do is demonstrate we understand how it all comes together. We know how to get to that endpoint.
I think showing with very modest investment, we've made a huge leap on the chemistry side that gives us a very clear path. It's not just a vision of a technology roadmap. It's actionable. We can deliver on these things. Proteus is the big leap we needed. It becomes much more of a sort of an architecture of the consumable, how many spots do you put, how big, intersecting with the chemistry. Maybe with that, the other part of the chemistry is not just cutting. It's how we detect the amino acids. Let's have John come up and give you a dive into this world, pretty fascinating world of data generation and how you do the development of these recognizers.
All right. Thank you. I'm John Biacelli. I'm the Chief Product Officer at Quantum-Si. I'm going to give you an update on our progress towards detecting all 20 amino acids on our sequencing platform. Obviously, a longstanding goal and vision of ours is to enable the detection of all 20 amino acids and develop end-terminal amino acid recognizers, which are these protein binders with labeled dyes to enable the detection of all 20. As Todd showed, what Proteus enables now or delivers is a detection system coupled with these new dyes and advances that we're making using AI for binder and protein design and protein structure prediction. We're in a position now to deliver on this vision. I'll go through the details of how we're going to do that. Let me lay out the roadmap for how we're going to get to 20.
I'll give you where we are today and then how we're going to get to detecting all 20 amino acids on the platform. If you're not terribly familiar, Todd gave a little bit of an overview on how the technology works, but I just wanted to give an overview for anyone who maybe is less familiar. Essentially, we take proteins, we digest them into peptides, we immobilize those peptides on the surface of our Kinetic Array on the Proteus platform. We then introduce the end-terminal amino acid recognizers, which again are just our dye-labeled proteins that bind to the end-terminus of those peptides. We're measuring the fluorescence signal of those binding events, so they're coming on and off. There are also amino peptidases, which are present in the solution. Those periodically cleave off the end-terminal amino acid, exposing a next amino acid in the sequence for detection.
Here is an example of RLIF being sequenced on the system. You see the detection, cleavage, followed by detection of the next amino acid. That leads to the plot on the top right of the slide. That gives us a trace of those pulsing events. We are able to color those pulses by the intensity and color ratio. That gives us the recognizer identity. Certain recognizers do detect more than one amino acid. We then get the amino acid identity through the kinetics of those binding events. We take those on-off events, and from those, we can determine what amino acid was actually present there. That is how we go from the fluorescence signal to the actual amino acid sequence of the peptide that was on that surface. Since the launch of Platinum in 2023, we have been iteratively making updates to the sequencing kit.
With each one of those kits, we either bring out new recognition capability or we improve on existing recognizers. The most recent release was in Q3 of this year. We came out with Sequencing Kit V4. That added recognition for glycine onto the AAS binder, bringing the total number of recognized amino acids on the platform to 14 using the six recognizers that I have shown on the slide. One other way that is a little bit tangential, but I wanted to highlight that we have at our disposal, and Todd kind of discussed it as well, is just in the software and the analysis and improvements that we can make that also enable detection of more amino acids. We have been working on that, making signal processing improvements. We have been able to show that we can extract methionine binding events from our NQ binder.
We've now added methionine to the NQ binder. I showed a couple of examples of peptides that we run internally regularly on proteins with those methionine binding events. That, of course, now brings us to 15. We have an update that's coming out very shortly. We're working on the final stages of a Library Preparation Kit, V3. With that, there will be a software release which will enable the detection of methionine and bring us to 15 on the system very shortly. How do we get to 20, though? We're now at 15. This sort of lays out the roadmap for how we're going to do that. Things may vary slightly from what I've shown here, but this is our internal data as of today and how we think we're going to get to detecting 20 ultimately.
Let me share sort of where we are and what we have internally and show you how we're going to get to 20 amino acids. That will be done most likely with eight recognizers that I have shown with the groupings of the amino acids on the slide. We already have a version of the R binder that has histidine and lysine binding. That is shown in the traces there on the left in the center of the slide. That binder is in development. We're currently working on that, and we have data internally with those binding events. We have a new proline binder. That is the one shown on the right. That is a new binder that we've been developing, and I'll talk about that a little bit later on in the presentation.
We have been using AI techniques for protein binder design and structure prediction and been able to develop this proline binder. That leaves two remaining amino acids that we have to get to, which are cysteine and threonine. We have campaigns underway internally for the development of those. Most likely, the cysteine from preliminary data we've seen internally will be on the NQM binder. We will be working on the development of a new binder for the detection of threonine, which we think will also bring along serine on detection of that. This is our pathway to getting to 20. We're accelerating our recognizer development pipeline. I'll go through that acceleration. We're going to heavily leverage AI and protein structure prediction methodologies to accelerate that development. We're scaling everything up.
The goal is at Proteus launch in 2026 to be detecting 18 amino acids on the platform and then bring the full 20 to the platform in 2027. We are going to have internal demonstration of that 20 in next year and then bring that to the platform in 2027. That is our roadmap for getting to 20 amino acids. As Todd showed, sorry, the detection system on Proteus has already been de-risked. We have the eight dyes. They have been demonstrated on the Proteus prototype systems internally. We have additional space if necessary. If we need to make any changes and go to additional recognizers for any reason, there is space within this two-dimensional plot to add additional dyes if necessary. We have the detection system.
We have the dyes, and we have everything to support that roadmap that I showed on the previous slide using eight recognizers on the platform. Let me go through the pipeline and sort of how we do this development and how we are accelerating it and how we think we've scaled up and we're going to be able to achieve that vision of delivering all 20 amino acids on the platform. I wanted to show this as sort of the evolutionary tree of recognizers that we have on the system and sort of how do we get these? Where do they come from? How do we evolve them? This shows the current set of six that we have in the Sequencing Kit V4. There's a couple of different ways that this happens, right?
If you look at the top three, they sort of follow this linear trajectory. We're making improvements to each one of those. With each improvement, we get better coverage. We see more sequence context with that recognizer. We are constantly delivering improvements on those. We continue to do that, always looking to improve the coverage of existing recognizers as well. The other family that's at the bottom is a different pathway where we had this NQ binder. With that, we're able to evolve, keeping the same backbone, evolve the binding pocket to get recognition of new amino acids using that same backbone. In this case, we were able to get the GAS and the DE binder by evolving the NQ binder to do those additional recognition events. This is all protected by 27 different granted patents and applications.
We obviously have a lot of expertise and long history of doing this and capability that we've developed over time. I also showed on here sort of how we've evolved these from the kit iterations. We had the V2 kit, which brought D recognition at the bottom there. V3, we updated four of the binders, so that gave us better recognition capability on some of those, as well as bringing DE binder. The V4 kit was finally bringing the GAS binder to the platform. Tons of experience on developing these recognizers. We have a lot of different ways that we can do that.
As part of all that development, we've also amassed what we believe to be the largest amount of data, both sequence and structural, for end-terminal amino acid binders that we can correlate to the kinetic data, to the binding affinity, and to the kinetics of these recognizers. We have tens of millions of binding data, often with different peptides, with different sequence contexts. We have structural data that we've mined from publicly available sources like the Protein Data Bank, which I'll show in the next slide. We also have X-ray crystallography capabilities. That gives us about 1,000 different structures of proteins with amino acids bound in the pockets of those. As we go through the pipeline, we obviously generate kinetic data as well. We have both bulk kinetic data that we've generated and single molecule kinetic data.
What all of this data enables us to do is feed it into AI models that can be refined to lead to the better prediction of new binders and also improvements to our existing binders. We have amassed a huge amount of data that we can use to further drive this process. This is just a quick example I wanted to go through. We have done things like taking the Protein Data Bank, searched it for all evidence of end-terminal amino acid binding. We have then refined that search looking for things that have a full atomic representation. They are exposed on the surface. They are just a single binding event. These are high-quality structures that we can then use both as scaffolds for new parents, like for threonine and proline, but also for model training to refine binders.
These are the kinds of approaches we're using to mine existing information and feed that into our pipeline to accelerate things. Of course, there's just been a huge amount of work in protein structure prediction and protein folding that's been done by both academia and industry. The Nobel Prize in 2024 was awarded for protein structure prediction and protein folding. Those have led to models that we are directly using, deep learning models that we're using for the development of our binders. Meta AI, which was originally the Facebook AI research team, developed an evolutionary scale model. This is a deep contextual language model that you can take data and refine the model to do prediction for end-terminal amino acid binders. That's how we're taking the volume of data that we have and feeding it into models to lead to better prediction for proteins.
Of course, this is all using NVIDIA GPU hardware acceleration. It's commodity hardware. We can scale it either internally. We can deploy it in AWS on the cloud, and we can execute all of this modeling using their hardware. We have not only our own data or our own expertise, but we have just a wealth of information that's coming from academia and industry that we can utilize to accelerate our pipeline. This is where we are. This is our pipeline. We have these different sources of data. We have combinatorial approaches. We have AI design. We have rational design, X-ray crystallography. We feed that all into our pipeline. We are scaling up the pipeline by about 4x relative to our historical amount of candidates that are going through. We've scaled up all the different methodologies we're using, utilize more AI techniques.
We are funneling that down as we go through this pipeline. We do screening. Then we get bulk kinetic measurements. We get single molecule kinetic measurements. That leads to binders that we then put into production as well as feeding back into new designs. This scale-up is what's going to enable us to deliver on the vision of 20 amino acids on the Proteus platform. I just wanted to end with sort of two examples of how we've utilized this information and how we've arrived at some of the binders and recognizers that we have on the system. This is one example where we actually use X-ray crystallography and then combine that with combinatorial screening and AI approaches to arrive at the GAS recognizer.
What's shown on the left is the actual NQ recognizer, which we were able to obtain a crystal structure with a G-terminating peptide, which obviously is not the natural binding, but under crystallography conditions, very high peptide concentrations, you can get this structure. What that structure does is gives us insights into how we would modify this NQ binder to be able to recognize a G-terminating peptide. We went through that, got some hits from combinatorial approaches, and were able to come up with the structure on the right-hand side, which basically filled that pocket, retained the G-amino acid binding capability, but also excluded the NQ binding capability. We can basically take this atomic representation.
We have atomic-level control over these recognizers and getting the recognition what we want while also getting the specificity and excluding the recognition that originally had of NQ and tailoring it exactly to GAS. That is one methodology where we took an existing scaffold, modified the binding site, and were able to get new recognition capability from it. The other example that I wanted to go through was the proline recognizer that we've been working on. This was, again, largely coming from AI techniques that we've been leveraging. We took a parent protein, put it through AI structure prediction models. That led to a hit. The interesting part of that hit is it has 18 different mutations, which, had we not used AI, would have taken us a very long time to arrive at using any other method rationally or combinatorially.
This sort of proves the power of AI and that we can get these hits very quickly, and they're very complex and things that would have taken us otherwise a very long time to achieve. One aspect that we are paying close attention to when we're doing screening is specificity. What these traces show is that while this was a good proline binder, we also got a little bit of isoleucine cross-binding from that. Of course, we want them to be very specific. We took the strategy with those 18 mutations and reverted each one one by one to the original parent amino acid. What the plot on the right shows is this is the binding ratio between the proline and the isoleucine.
One of those amino acids was largely responsible for the isoleucine binding, and we could knock that out by reverting that single position, still retaining the proline binding capability, but removing the cross-binding to the isoleucine. We have all these techniques. We have AI. We have combinatorial approaches. We have this whole pipeline that's all set up to enable us to deliver. Let me just summarize the path to 20 amino acids, our current recognizer pipeline. We've got the 14 in the V4 kit. We're going to go to 15 very soon. At Proteus launch in 2026, we'll be at 18 amino acid detection. We've scaled up everything by about 4x. That scale-up is what's going to enable us to demonstrate 20 internally in 2026 and then release that on the Proteus platform in 2027.
We're heavily leveraging AI tools as well as our own screening data to refine these models. We're using combinatorial and rational design approaches as well in concert with that. That is all going to accelerate the pace of recognizer development and deliver on the vision of 20 amino acids on the Proteus platform. I think that's it. Thank you for your time.
Thank you, John.
Let you slide out of there. Maybe just a couple of comments. I think the first one is you can't just start tomorrow and say, "I'm going to use off-the-shelf AI tools and recreate what we just did." The ability to accelerate now has a lot to do with everything we've learned over the last few years. Had I gone to the team two years ago and just said, "Go scale up 4x or 8x or 10x, just try to make this some sort of linear investment model where I tell you just double the people and double the output," you would have doubled the work and you would have doubled the output, but the quality and the rate of improvement might not have been any better. You could have created lots of bad candidates that really wouldn't have moved you forward. You would have spent more money.
You would have felt like you were being more productive, but you were not actually being more productive. By having all of this data now and all the knowledge that our team has, you are able to leverage that so that not only does the pace of the activity pick up, but you do that with still retaining a high quality of the candidate. It is really that intersection you need to have to speed up. If you think about what John and team intend to accomplish, we have been working on protein sequencing for many years as a company to get to 14, and we are going to close the remaining gap in the course of the next year. That acceleration comes from that data from all those learnings and from sort of the core competencies and workflows that we have built inside the company over the last few years.
Very excited about that rate of improvement and that progress. As that coverage gets to complete coverage, that sets up nicely for what Brian's going to talk about, which is what are all the things we can detect using this core technology? What's all the underlying rich kinetic data that's there, and how do you apply that to look at something like a post-translational modification that has huge implications in terms of biology research, biomarker development, and so on? With that, we'll let Brian come up and talk to you about that topic.
Thank you, Jeff. Today I want to walk you through our vision for delivering PTM applications on Proteus. I'll discuss some of the advantages of doing single molecule detection for PTMs, walk you through three approaches that we are developing to get very broad access to PTMs across the proteome, and then walk you through examples of how we're doing this in practice on the road towards Proteus launch. First of all, what are PTMs? PTMs are chemical modifications to the side chains of amino acids that can have a dramatic effect on protein function. There are an enormous variety of PTMs. It's estimated there are around 400 different types of PTMs in the body. In proteomics, though, there are really only a handful of PTMs that are of very high interest because they're either abundant or they're implicated in disease states.
These are things like phosphorylation, which is involved in cancer signaling pathways, glycosylation, which, for example, is involved in regulation of the immune system, and ubiquitination, which controls protein degradation and is important in neurological diseases. PTMs are really the determinants of protein function. In proteomics, researchers have come to understand that it is really not enough to look at which proteins are present in a sample. Researchers really need to know for the next generation of proteomics what PTMs are present in those proteins. Unfortunately, current methods in proteomics run into a lot of challenges when it comes to PTM detection. We can sort of categorize current approaches into two categories: mass spectrometry and affinity-based. With mass spectrometry, quantitation of PTMs, so what fraction of a given protein or peptide is modified, is a very difficult question to answer.
There's also the challenge of ambiguity in detecting the actual site that is modified. You might have the signal that tells you that a PTM is present, but pinning down where it's actually located can be difficult in many situations. Related to that, when there are multiple PTMs within the same peptide, that can create a really difficult situation for mass spec to deliver an answer on where those PTMs are. On top of those challenges, you have this workflow and analytical complexity with mass spec data, and that, for example, limits clinical adoption. Affinity-based platforms can have the ability, in some cases, to detect thousands of proteins, but they provide no PTM or proteoform information. Other techniques in the affinity-based space require they can't see PTMs, but they require site-specific and protein-specific affinity regions for every single protein, as Jeff pointed to earlier.
That results in this enormous proliferation of reagents and expense to access PTMs more broadly across the proteome. In contrast to that, what are the advantages for PTM detection using single molecule sequencing? The first is the way that we're sequencing proteins on Proteus, the nature of the assay is inherently quantitative. We're essentially counting single molecules. You get a quantitative readout. Another aspect of the sequencing approach is it's inherently site-specific, where we get sequential information and we detect a PTM, so we understand exactly where it is in the peptide. Related to that, that gives us access to more complex scenarios where there are combinations, which is quite common in the proteome, combinations of PTMs and potentially variants in the same peptide that might be impossible to approach with other techniques. It's much more straightforward with a single molecule sequencing approach.
Those advantages sort of sit on top of this platform that's really easy to use for customers. It doesn't have all this workflow complexity, and the data analysis is straightforward. Another really critical advantage is that it doesn't require site-specific and protein-specific reagents for every protein and PTM. As I'll walk you through, we have developed approaches that provide universal access to PTMs anywhere where they're located in the proteome. What are those methods that we're developing? There are three main approaches that we're developing that can be used either independently or in combination to achieve very broad access to many different types of PTMs no matter where they're located in the proteome. Those approaches are first using kinetics. I'll walk you through examples in subsequent slides of each of these. Using the kinetics that result from the presence of a PTM to detect it.
The second is pre-recognition. Here we use a universal PTM affinity reagent, which can be an antibody, that detects PTMs anywhere where they're located in any peptide, regardless of its length or sequence. The third method is direct detection. This is where we have a recognizer in the sequencing process itself, which could be one of our standard recognizers or one that we engineer specifically for a given PTM. It recognizes a PTM when it's at the N-terminus, just like our recognizers do for any other binder or any other amino acid. To walk through each of these methods in a little bit more detail, with kinetic detection of PTMs, the important thing to understand is that the kinetic response to a PTM is an inherent property of the recognizers that John discussed.
They interact with the N-terminal amino acid, but when they do that, they also make contacts with downstream amino acids, and those contacts inevitably result in changes in the kinetics of that interaction. It's kind of a universal biophysical phenomenon that we see all the time in our assay. What that means is that many different types of PTMs result in these detectable kinetic changes. Examples of those would be things like phosphorylation, methylation, oxidation, and many others. Our customers are actually using kinetic detection now, in some cases, to look at peptides that have these complex combinations of PTMs, and they can see which PTMs are located in these complex arrangements just based on kinetic data. With pre-recognition, the important thing here is that using a labeled affinity reagent, we can detect a PTM anywhere in a peptide. This allows for a few kind of interesting benefits.
The first is we can combine affinity reagents for different PTMs in the same mix and do a pre-recognition step where we can detect multiple different types of PTMs in the same peptides in parallel. A great example of that is a method we're developing that we call the pan-phospho-detection, where we can recognize all three of the major types of phosphorylations in the proteome, phosphoserine, threonine, and tyrosine, in one assay. What happens here is the proteins are loaded on the chip, and for a short period of time, the peptides are exposed to this affinity reagent that binds to the PTM. That reagent is washed out, or that combination of reagents is washed out, and then we sequence the peptides. This method also has the benefit of being extremely sensitive to stoichiometry, which is something that other techniques really struggle with.
That results from just the nature of the output of the assay. It's a very, very clear pulsing pattern when we see on-and-off binding of these affinity reagents to the PTM. For N-terminal detection, the important thing here is that this doesn't involve any change to the workflow. It doesn't even require a pre-recognition step. This is simply sequencing the peptides in the presence of a binder that binds to a PTM when it's at the N-terminus. Like I said, those could be, I'll actually show you an example, those could be standard recognizers that are already in the kit that also bind to naturally or unmodified amino acids, or they can be engineered for specific types of PTMs. This has an advantage of being really useful for de novo sequencing applications because we're getting that direct N-terminal detection.
Now I'd like to walk you through some examples of using these techniques in practice. The first is kinetics. Here we have two peptides. One has a tyrosine at position three, and the other one has a phosphotyrosine. What we see here is changes in the pulse duration, which is one of the key kinetic parameters of the recognition segments leading up to the position of the PTM. The presence of the PTM is changing the way that the binders bind on and off to the peptide on the way of the reaction getting to the location of the PTM. For example, here, you can see the pulse duration of the arginine decreases from 1.3 to 0.5 seconds. We see other kinetic changes. We actually, in collaboration with a lab at University of Virginia, published a paper where we use this technique to detect triple-myosin proteoforms.
To give you an example of just how broad this approach is to different types of PTMs, here's an example where we use the same technique for detecting citrulline. Citrulline is a PTM of arginine, and it's very difficult to detect by mass spec because there's a very small mass difference between the modified and unmodified forms. Here, where we have an arginine at position four in the peptide or citrulline at position four, we get these sort of dramatic changes in pulse durations at multiple upstream positions that tell us where that PTM is located. Actually, the information is much richer than that. It goes beyond pulse duration.
Just to give you an example of one position in this particular peptide, the leucine before the citrulline shows changes not just in pulse duration, but also interpulse duration, which is the time between the binding events of the leucine recognizer to the peptide. It also shows changes in the RS duration, the recognition segment duration, that essentially measures the rate of cleavage of the amino acid. These multiple kinetic parameters are changing at multiple upstream positions in a very straightforward and predictable way, and our software can really leverage that rich source of information for accurate and sensitive and very broad PTM coverage. To switch to pre-recognition, here's an example of using pre-recognition that demonstrates how this technique can be applied to PTMs that are located anywhere in a peptide and in any sequence.
We've taken here multiple synthetic peptides that have phosphotyrosine at known positions, either at the N-terminus, at internal locations within the peptide, or just adjacent to the C-terminus, which is where we link the peptide to the chip. In this case, it's an antibody that recognizes universally phosphotyrosine, and we do this pre-recognition step for these peptides simultaneously on the same chip, and we're able to detect phosphotyrosine with this very high sensitivity no matter where the PTM is located. We took that technique and recently applied it in a collaboration with a company called Karna Biosciences, where we're probing a panel of important kinases. With kinases, tyrosine phosphorylation status is a really important indicator of protein activity, and these proteins are used in drug development assays, so that's quite important. Here, what we've done is I'm just showing one example, protein tyrosine kinase 2B.
We take this kinase, digest it as we normally do for our proteins, load the peptides on the chip, and expose them to the PY antibody in this case for 30 minutes. This allows us not just to detect tyrosine phosphorylation at multiple peptide positions, but also to sort of pinpoint the location and determine quantitatively what fraction of those sites are actually modified. That is really difficult information to get with other techniques. We can take that sort of mapping and frequency and map it onto the full-length structure of the protein. Finally, to give you an example of direct N-terminal PTM detection, what we found actually with the NQ recognizer that John was talking about, that now with software improvements and things like that, is recognizing methionine. We found that it also directly recognizes oxidized methionine, which is also called methionine sulfoxide.
That is an important PTM in multiple diseases. What you can see here is we have a protein called IKBA that we've digested, loaded on the chip, and in one of the peptides, there's a methionine at position four. What we see are two sort of different phenotypes of kinetic signatures for this peptide. One has really short pulse duration, but still detectable for the methionine at position four coming from that NQM recognizer. The other one has this dramatic shift, the increase in pulse duration. That is due to the impact of that chemical modification to the methionine when it's oxidized. We have very clear, direct N-terminal recognition of a PTM with a recognizer that we already have. Of course, like I said, this could be applied with other types of recognizers that we can use our pipeline to develop.
To summarize, our path to broad PTM coverage on proteome really involves four key areas of development. The first is recognizer development, which John went over quite thoroughly. It's very important for PTM applications because ensuring that we can see all 20 amino acids at every position, that is what gives us the information to detect the PTMs using these approaches that I outlined. The second is higher output on Proteus and also combined with longer read length. That's going to enable deeper proteome coverage and more sensitive PTM detection because we're getting more reads, and we're going to be able to look at more proteins than we currently can on Platinum. Another development area is on the software side.
As I've indicated, we're really getting a super rich source of information from the kinetics when PTMs are present, and it's really ideal to take that data and train AI models that will allow us to create software that can pinpoint and automate the detection of these post-translational modifications. Finally, of course, we're continuing development on these three methods: kinetics, pre-recognition, direct detection to achieve the broadest PTM coverage across the proteome. Thank you.
Thank you, Brian. As we move into my presentation, I think one maybe point to make there is that competency we've built internally, how do you take massive amounts of data, use AI to train and improve our recognizer pipeline? You can sort of see the thought process that Brian's bringing to the table with now how do we do the same thing and mine the richness of the sequencing data we generate to uncover more and more things to detect these PTMs. Some of these core competencies that get built for internal purposes can translate over time into applying those to our product to improve its capabilities. I want to talk a little bit about what have we learned in the commercialization of our first-generation Platinum and Platinum Pro.
How are we taking those learnings and thinking about and applying it to the requirements for Proteus, and I think tie it to what you're hearing today around where our focus is with the chemistry and with PTMs. The first thing is by having this technology in the market, we have been making improvements, and every time we make improvements to the technology, we learn about a new application, we learn about a new thing a customer wants to do. Those improvements in amino acid coverage and lowering the sample input as we expect to do with the new Library Preparation Kit, all of those things port over directly into Proteus. None of that work is lost. It's all directly portable.
You then intersect it with that higher sequencing output, with the ability to read deeper into the peptide, and with a consumable that has just a significantly more favorable cost of goods, and you really get to that sort of inflection point in terms of the total technology picture. It is also commercially, how do we think about going to market? With Platinum, we went to market as sort of a general technology. We said, "We've got this protein sequencing technology. It can do the following things. How would you like to apply it with customers?" Now, over time, you've seen us do some application-specific work. Barcoding is a good example. We were in biopharma. They started to work on this peptide barcoding application, and we applied that and created a kit to sort of make that more efficient. We were really sample-prep agnostic.
It's thousands and thousands of different reagents people use to prep samples in proteomics. I think if you've come from the DNA world, you underestimate how complex the sample prep is in proteomics. In DNA, you can pull down your DNA, your RNA, or all of the nucleic acid, and then the magic of PCR to sort of level out the playing field, bring up the low-abundant things and see them. That doesn't exist here in the proteomics world. We tried to put that aside. We tried to let that be more of a thing the customer defined and then would flow into our platform. While that can work, it takes a lot more work to onboard a customer. It takes longer to scale that up.
If you want to go after really difficult, valuable things like super low-abundant biomarkers, integrating those things is a far better approach to bring products to market. I mean, we market it sort of to all comers. That's pretty common. Go out there, talk to researchers, see who's sort of interacting with you, who has an interest in the technology. That works in the early stages. That works with those early adopters. I think we've learned a lot about really what segments to target, how to focus to get a depth of application to really drive the platform purchase and then the eventual pull-through of reagents. The other advantage of being in the market is when you're not in the market, you sort of rely upon classic tools, right? You do various forms of market research, voice of customer.
These are sort of the phrases we like to use as marketers. When you're in the market with a commercial product, like we are, we have a scientific affairs team, Lean and Mean Group, but they did something recently called the Platinum Pioneer Grant. This is us actually just working with customers to understand what are you trying to accomplish, what capabilities does it require. This is really meant for us with our current platform to identify people who would get a certain amount of reagents to do some work with and publish their results. This is part of the market development activity we do with the Platinum Pro. You can mine this and see what I talked about at the beginning is what's happening. The complexity of the types of analysis tools customers need shows up in this data.
More than 50 researchers submitted detailed proposals. When I say detailed, I mean they told us, "I'm trying to study the following thing. I need to detect the following types of things in this sample, in these proteins or these PTMs." When you mine those, you see two things that tie back to where we started. One is more than 50% of the applications need three or more types of protein analysis, meaning they can't complete their work by just doing protein ID or just looking at a PTM or just looking at a pulldown protein versus maybe a mixture. They need multiple sets of capabilities. When you get into PTMs, if you look at the proposals that wanted to work on PTMs, which is a very large number of that 50, 65% more need two or more.
That means very few applications to solve difficult biological questions are as simple as just detect phosphorylation or just detect methylation. They need to sort of tackle these problems that are very complex and be able to do it. Again, absent a platform that can do all of this, it leads us to where we started, which is people try to then own multiple different platforms. That means a lot of the work just stays in the core lab. We know this data from our customers. We also know the word academic medical center. We use these phrases a lot as we all interact. We're selling to academic centers. We're selling to biopharma. We're selling to industrial. I could do this type of a map for every single one of those segments.
This is something, again, very difficult to tease out when you're doing market research, very easy to tease out when you have people going in and calling on these institutes every single day. I'm just taking one example again to do this across other segments, but let's take an academic medical center. They've got research labs doing very basic biology research. This could be very fundamental work. It could be even single molecule, just biophysical type of work. They have a huge range of application interests. They're typically very low infrastructure. Most of these folks send to the core lab, and they tend to be lower utilizers. The labs are smaller, less staff. Their work sort of ebbs and flows. That core lab, that's exactly where we started today talking about.
Those are the folks who have the luxury of having the infrastructure and the dollars to own lots of different pieces of equipment. The publication that Brian talked about, some of the other ones that have entered the market from people using our tech, is often them using our Platinum Pro as a complement in the core lab to other technologies. Translational labs, as we get more, why are we pushing towards PTMs? Why are we pushing in that direction? As Brian described, that's where the functional part of biology is happening. That's where the disease biomarkers, the therapeutic responses, the risk of recurrence, these types of analysis or studies people want to run, that's really in that translational lab. What do we mean by that? That's somebody who's got a foot in research and a foot in clinical practice. You have clinical labs.
These are folks who are just doing routine testing. I think when people think proteomics, they often think, "Oh, immunoassays or that tau biomarker they're going to run." There's also a long list of what get called esoteric tests. I showed you sickle cell disease. There's a list of more than 25 different hemoglobinopathies that today get done clinically to treat people, and they're cobbling this together using technologies like capillary electrophoresis or HPLC. Overlay that with Platinum based on the capabilities of that platform. When I say capabilities, I mean what's the amino acid coverage, what's the level of automation, what's the throughput, what's the sequencing output, all these things we've been talking about with Proteus. What do you sort of see, right? A very low-cost platform, pretty simple workflow, automated data analysis. It had a draw into basic research.
What were the reasons, right? It could solve their technology problem. It had a low capital cost, so they could take control over their research. They did not have to send to a core lab. They were able to afford to do this on their own, implement the work. That is great in terms of a proof point of if people could access the technology. We think about Proteus. We do not want to be a million-dollar piece of equipment because there are all these other labs who would want to insource. These folks were a good beachhead for us, but they are lower utilizers, right? They buy, they run the kits, they publish, and then not until they start their next study. Core labs, we have been largely a reflex testing technology, right? One of the papers from Northwestern, they were doing work on phosphorylation.
They used our tech to resolve ambiguous cases to try to reduce the false positive, false negative rate. An important tool, but we were sort of relegated into more of a reflex environment. Similarly, on the translational side, Brian mentioned the study in osteoporosis with University of Virginia, again, getting used where we had the capability, we found access into the translational labs. The clinical lab, really not accessible with the current platform, both mainly from the standpoint of sort of automation and just total sort of end-to-end workflows. People do not do the heavy lift. I spent a long time in clinical before coming here. You have to provide a very buttoned-up product from sample all the way to report to really go into those environments. They do not have sort of the developmental capabilities that some of these other labs have.
Now take those learnings that we've been getting in the field and overlay it on what you heard today. Why move to a nanowell array? Why scale up the sequencing output? Why have more samples per run? Why have this timeframe? How do we get more chips run through the machine in a period of time? It comes down to these points, right? More samples per run with more output. We know that in those translational labs and some of those core labs, we're going to have to scale those attributes of our technology to move from sort of a reflexive position to a primary position.
Building upon what Brian talked about, some people might want to run a lot of samples with a little bit of information, but others might want to do very deep analysis, a complex mixture, look for a very rare event like a variant or a PTM. Having a lot more sequencing output enables that. That is something you are limited by in our current platform with only two million wells on the chip. Multiplexing, as we think about maybe lower-cost applications, basic protein identification, the ability to take advantage of all this real estate, combine it with multiplexing technologies to really lower the cost per sample in those more cost-sensitive applications. Something feasible to do with both the output and the cost structure of the protease chip, something we cannot really do effectively based on the output and cogs of our current chip.
It opens up some sort of commercial opportunities or pricing opportunities to access certain markets that might be complementary to sort of the core desire to be focused more on PTMs and variants. Again, tie out to what we heard. Why get to 20 amino acids as fast as possible and focus on PTMs? If you dive into that data, dive into those pioneer grants, this is what people want to do, right? They want to look at PTM profiling. They want to look at multi-PTM profiling, especially in those core labs and the translational labs. Single amino acid variants, interesting opportunity both in translational, but also potentially into the clinical space. Sequencing of variable regions of antibodies. You look at some of our biopharma customers, some of our industrial customers. They'd love to be able to do that, but that requires that expanded coverage.
That requires that depth of sequencing to really look at those variable regions, something not accessible today with Platinum and Platinum Pro, but will be feasible as we move into Proteus. Biothreat detection surveillance. We talked about on our last earnings call that our instruments are deployed in the Department of Defense today. We think that work that's going on there, we have visibility to it. We know that they're going to publish some data. I assume they won't publish everything, but there's an opportunity there to continue to scale up the capabilities that that market needs to really think about what's the future of biothreat detection look like. We think the sequencing coverage, the output, the automation, all these things play well into where we believe they might be going.
You never get told the full picture, but we believe those attributes will sort of dovetail nicely. Automation matters. I like to put things in sort of numbers. If I wanted to take a Platinum Pro today as a customer and I wanted to run four samples, I get the kits, I unbox them, get down to the nitty-gritty of this, right? Someone would have to handle about 30 different reagent tubes to process those four samples. If you go into a lab doing mass spec, doing other things, they would not sort of flinch at this. They are used to handling lots of reagents. They are used to doing a lot of pipetting. For our product, we are acknowledging here, it is over 100 individual steps to process a sample. Again, in those basic biology labs, there are people doing gels. There are people running HPLC.
They're taking samples over to a mass spec core lab. They're cobbling together lots of different things to do their research. As we were thinking about the scale-up in the output, as we were thinking about the coverage in PTMs and the markets it might unlock, we wanted to automate. This is really the reason why. Flash forward to Proteus, those same four samples, now you just take this reagent tray and you load your four samples and you stick it in the machine, you add the chip, you add a tip box, you hit a button, it's over. There's no more pipetting. There's no reagent formulating. There's no mixing. All of that's happening in the machine. The level of automation of the lab workflow is significant with the Proteus platform.
The other thing I mentioned in the beginning that ties through is really applications, the sample prep and enrichment, and also the application of AI, not just to our internal recognizer development program, but also to the analysis of our data. For any of the people who followed genomics, in the early years of the revolution of NGS, it was largely software provided by the company. Over time, that opened up, and there were lots of companies who built analytic tools on top of that sequencing data. I suspect the same type of thing could happen here. We also are thinking about it in terms of what are some of those clinical applications. Could we marry up with someone who's doing the enrichment that we could marry with the capability of our sequencing and deliver a validated workflow and go tackle some of those challenges?
We have been building up over the last few months sort of a pipeline of potential partnerships across all three of these areas and wanted to today sort of provide insight into a couple of those. We expect to provide more information coming forward here on the other ones that are in the pipeline. Here are a couple of examples. Brian mentioned the first one, Karna Biosciences. This is an interesting application of our technology into much more of an industrial setting. These folks are a leading provider of assay-grade kinase proteins into biopharma drug discovery. They are looking at our technology as a way to evaluate and sort of validate the phosphorylation profile of these proteins. This could potentially be useful for their customers to have this information and have it in a quantitative way when they are buying products.
This is an opportunity for us to—this isn't necessarily—we're not selling Karna Biosciences' products, but can they leverage our technology? Can they improve or in any way enhance the offering to their customers? Depending on how the data shapes up and the collaboration, could we work together with sort of complementary reach to bring sort of these types of solutions into the market? Flipping to the right-hand side with SiennaQuant, people who follow mass spec probably recognize the name SysCapa. That is a company that has made enrichment for mass spec for many, many years. It is widely used as an enrichment technology in mass spec. They have created another entity within that broader holding company focused very exclusively on ultra-low abundance biomarkers. Think your cytokines, think your interleukins, things that are historically very, very difficult to detect and very unreliable to measure quantitatively.
They're at such a low abundance. They've been developing some interesting technology to enrich for those that we're now working with them to combine, do their enrichment folded into our library prep and ultimately into sequencing. Now you can start thinking about, is this a way to create a very accurate low-abundant biomarker panel? There's a lot of clinically relevant biomarkers that are at these levels. This could be a very interesting way for us to couple a technology, validate that whole workflow, and bring it to the market. Rather than going and saying, "Hey, sequence these things, and you need to figure out, Mr. or Mrs. Customer, what to do on the upfront," this brings it all together and brings that buttoned-up package as we think about, again, moving beyond basic research and core and into translational and clinical.
We think these types of partnerships can enable very interesting analysis in terms of what the data will likely prove, but also creates a way to actually reasonably implement it. It eases that sort of onboarding process. Again, sticking with academic, again, we could do this for other segments of the market, but let's focus here for a minute on academic. How do we see everything you've heard about today with Proteus play out compared to what you saw a few minutes ago in terms of Platinum and Platinum Pro? We think the basic research lab will still want to access this technology. Now, we don't intend to price this instrument at the $100,000 sort of price point that Platinum is at. Some of these folks might not be able to insource, but they might want to send to the core lab who has that.
The opportunity to continue to cultivate that volume and ensure it ends up on our technology within a given institute is something we'll keep a focus on. Some of the larger basic research labs will certainly have the capacity to purchase, but some smaller ones might not. The core lab really stepping out of that reflex tech or that I'm solving for the edge case problem you can't do and moving into more of the high-value applications. The ability to do broad PTM profiling is a clear opportunity in these laboratories. Being able to do increased throughput is important. One thing about a core lab is they tend to have batches of samples. Being able to do higher throughput, larger quantity helps fit. We can really move more towards a frontline platform instead of being that secondary or reflexive platform.
You put all this together and some of these ideas around partnerships with workflow validation and sort of end-to-end sample to report, and we can make very meaningful progress into translational labs and we believe even into clinical labs. I think this time last year we thought clinical could be five years or more out. I think with the right partnerships and the right validated workflows, this is something that can pull forward in terms of our time to access these markets with Proteus. I do not think it is a five-year thing. I think this can be pulled back in time. Again, are we going to take on the single protein biomarkers? That is not what we will focus. We will focus more on those complex esoteric tests that have high value but are difficult to do today and are underserved.
Again, an opportunity to be in another type of lab who's a consistent utilizer, consistent run rate, so that in total you get sort of the level of depth you need to really drive the adoption of a higher-priced platform into this market segment. Again, we've done these types of maps over every segment that we're in today with the Platinum and Platinum Pro system and really understand how many call points do we have and what do each of those call points need in terms of capabilities. Again, playing off of that, we're in academic research today. We're in some pharma biotech, defense, very recently into agricultural. I think the commercial or industrial sort of fits with sort of where the Karna Biosciences side of the world is.
We really see this opportunity to take that approach, that deeper profile of each account and apply it over top of each of these segments. The one thing that we've learned in each of these segments is the ability to see PTMs, to sequence deeply. These are the capabilities that unify all of these. People are looking for, in the translational space, the presence of a PTM may be indicative of a disease or not, or treatment response or not. In the same way, in an industrial setting, that can also show a process that's out of control. Someone's endpoint is different, but the capability they need is very similar. We think we've brought that together and have the right requirements in there that we can take that approach.
We just walk through with academic centers and translate that out into other segments of the market. What's that mean? How do we get from here to the end of 2026? Obviously, huge milestone to have announced that we sequenced on Proteus, on that prototype, and Todd took you through that data. We feel very good about that top line you see running across and getting us to that launch. Those fully integrated systems delivered and operational. That's that fleet of machines we'll use that will really allow Todd to take that data that you saw and the teams to take that data and now optimize the sequencing chemistries, optimize the workflows and performance to get even better performance than you saw today. Again, we're already exceeding the current platform, but we expect to continue to improve upon that.
Get out into Q2, we should be doing end-to-end sequencing. Those fully integrated machines should be seeing samples coming in and us getting sequencing results coming out of those machines. That sets us up then to work with some external customer collaborators. We're calling it customer collaborators because the point of this is not necessarily to sell these people an instrument. The point of this is to put an instrument in a select number of very sophisticated places who have the other methods available and do some comparison, really put some mileage on this technology, make sure the results we demonstrate internally get demonstrated in somebody else's hands.
We want to run that a little earlier than you might run sort of a classic early access program because in the event we learn something, we want to be able to fold that into this program and make sure we accommodate for anything we learn before we get out into Q4 and execute on the internal validation studies and the launch. Before hitting commercial ones, I'll bring your attention right below the launch one out there in the fourth quarter of next year, building off of what John said. We expect that we will demonstrate that we're capable of detecting all 20 amino acids by the end of next year. Again, we expect 18 of those will be in the launch kit.
With this demonstration in hand by the end of next year, we'll be able to roll through one upgrade to the chemistry to get to all 20, and we'll be able to do that in 2027. A pretty significant leap from 14 today to that point, either by launch or within that first year of launch. Commercially, I put up the milestones. I'll tell you that we may move these around, meaning we may pull some of these back over what's here, but this is our thinking right now. The reason we might pull some of these back is as we're in the market and customers are starting to hear more, we're already getting questions about what's the list price going to be because I'd like to think about how I budget for it.
Right now, we've pegged to release that list price in the second quarter. We may decide to pull that forward a little bit for those customers who are trying to do budgetary quotes, but we think that's an important thing to release both to set sort of the expectation in the market, but also to give people sufficient time to work through budgets, work through if they need to file for a grant or anything like that. When we get out into the third quarter and we've done some of the work with those external customer collaborators, we expect to share a more detailed update on what capabilities are going to be there at launch. Which of the PTMs will we be capable of? What's sort of the coverage looking like, the depth?
These items that Todd was talking about will start to provide some insights into what do we think that's going to look like at launch, again, helping build for the customer into that launch. We have an installed base of users. We have people using our system now under the placement program. Those folks are going to want to see an upgrade program. We hope that we can upgrade many of those people, so we'd expect to release that upgrade program in the third quarter. Again, might pull that forward depending on what we hear from customers as we continue to move throughout the year. I'm sure this event will spur some interest and might spur some of these conversations. I see that as a positive if we need to pull some of those events forward and give that to customers.
If they're thinking about how they move to protease, we think that's a net positive in terms of our long-term success. We wouldn't be disappointed by having to do that, and we'll just sort of listen to the market and make those sort of moves as we need to. Really just to tie out, I think what we really wanted to demonstrate today is protease, not only did we make a leap and that technology works, but we made that leap and have that vision and that roadmap that's very actionable to deliver on the long-term game here, which is eventually people are going to want that billions of reads and de novo sequence samples. This isn't just a vision. This is something we've actioned, something we've been executing on, and we believe the protease architecture is the starting point of that, but intersects with these other items.
We're going to continue to use Platinum Pro as a market development tool. Again, that doesn't make the revenue line today jump, but as we talked about on our last call, right, we were able to put 12 placements into the market in just a couple of months and over half of those in academia. These are the folks who get engaged, get using the tech that then are the ones who are going to want to know about the list price and the upgrade program and these types of things. We think that's a tool not only to get data into the market, but very useful as we think about building the early sort of set of customers who would want to buy a Proteus machine in that first year of launch. We got to execute. There is a lot of work to do.
We think that optimization work is going to only improve performance, but we have to go do it, and we have to finish that work. I think the plan is there. The team, you can be the judge of that for yourself. This is the three key leaders in the group. I can tell you that the depth of talent below these folks across all the disciplines would sort of surprise you to the upside of how deep we are and how talented we are across the board. I have high confidence that they'll do that, but we would be remiss if we did not focus on that execution. That partner ecosystem, we talked just about a couple today. We have things in flight with multiple other groups.
Again, really anything to do to accelerate application development, deliver sample-to-report workflows that may allow us to access some of those translational or clinical opportunities. We will keep that modest but targeted investment going on the technology sort of development pipeline, right? We want to bring Proteus to market. We want to build upon its capabilities that Todd talked about with faster sequencing and with getting more throughput from that device, but we also do not want to lose sight of having the next intersection point, that next leap. We want to be the ones always making the big leap on technology ourselves, not waiting for a competitor or somebody else to make a move. We will continue to keep that lead. We will stay focused primarily on development programs, but keep that sort of targeted investment in the background on the roadmap.
With that, we'll stop there, and we're glad to take questions. I think we—I forgot to bring the mics. I apologize. Just for the webcast, if you can ask the question into the microphone, that will help us.
Thanks. Sorry. Kyle Mikson Canaccord Genuity . Thanks for the day. This was a good presentation. On the pricing of the box, you announced that kind of early next year. You were saying, I mean, I guess it's going to be more than Platinum, but either way, the consumables are important as well. When you think about price per amino acid, price per peptide, how do you sort of think about that? What do customers—how do customers think about that, and what do they want?
And then also, kind of on that note, the super Poisson loading relative to the 30-40% single molecules filled into a well, basically mobilized into a well, how much further can you take that? Can it be 60, 70%, or could it be closer to 100%? Thanks.
Okay. I'll take the first one, and I'm going to—Todd, start walking up here so you can help on super Poisson. I don't want to mess up your math. First of all, on the list price front and the consumables, so yeah, we're obviously not going to be at $100,000. We think $1 million for a piece of equipment like this is just sort of way too high. It would sort of limit the number of people we can sell to.
If I had to give you sort of a general range finder, we think it's probably in that $300,000-$500,000 is where the machine's likely to fall. We're going to refine that as we're working with customers, and we'll have a more definitive position on that, but we think it's going to fall more in that range. In terms of consumables, let's start with what we've seen and then how that informs what we think about. When we are tackling—when we're trying to do an analysis that people have other ways to do, we see more pressure on pricing today. Pricing today is about $500 a sample. If we're trying to do something that they have other technologies that can do, we see more pricing.
When we're solving a problem like the publication Brian talked about where we're looking at an isoform they can't do with mass spec, or we're resolving the ambiguous PTM results from a mass spec run, we don't see really any pressure on that pricing. I think as we're thinking about consumable pricing, we're thinking a lot about, might we have application-specific pricing or do other things to sort of just try to mirror the customer in terms of when are we offering that value and the customer's comfortable paying for it versus when we might have to be more price-sensitive and compete with other tech. We haven't really decided the model exactly yet, but that's what we see in the market today. The more we skew towards PTMs, high-value biomarker panels, the more I would expect that our ASPs will be able to be retained.
Either way, the other component not to lose sight of in the gross margin calculation is the cost of making this consumable is going to be significantly lower than the cost of making our current one. We can have quite a bit of price sort of range finding by application and have stronger margins than we can achieve today with our current tech. Maybe on super Poisson, where can you sort of get to with that, Todd? We're about 33% single molecule loaded today. What would it look like in a sort of optimized world?
Yeah. I don't really have a definitive answer today. It is still an active advanced technology development. I don't really have data to share today. I know from past experience and other systems, we were able to push that pretty high. We were able to push it above 90%.
This is a single molecule world. Things tend to be a little bit more challenging. Proteins are chemically diverse. That might introduce additional challenges. My expectation would be that when we initially come out with it, it could be getting to that 60% range. And then over time, we would continue to improve it with additional kit releases and improvements to the workflow. I expect that over time, we'll be able to get into that 90% range, but it may not come out all at once like that. It may take a few iterations to get there.
Yeah. I think while we hand the mic for the next question, I see a hand up. Do you have a follow-up there, Kyle? Go ahead.
I did. Please. So on the PTM note for either of you guys, I think 400 is what you mentioned in the body.
There's a few that are relevant, though. How many will you kind of launch, I guess, launch with those capabilities? What's the timing? I don't think it was on that slide with the timeline of stuff.
Yeah. We haven't decided exactly what we'll launch with. What we did talk about on the slide, I'll see if I can reverse it here. The launch capabilities we'll have in Q3. We'll lay out sort of what will the device be capable of. This is going to cover not only the types of PTMs, but it also will give more insight into will there be library prep improvements and other things that get married in. We'll sort of share a comprehensive sort of launch capabilities update in that Q3 timeframe. You are correct. There's a long, deep list of PTMs.
That list probably shortens down to 20 or less when you get into the sort of the really biologically important ones. We're out right now doing work with our existing customers and some prospects of Proteus to really understand how to prioritize those. Obviously, phosphorylation. We showed a lot of data here. That's high on our list and something we would expect to have. Which of the other ones we tackle is really subject to some of this work we're doing, trying to, again, find the ones that are important, actionable in terms of their biological question, and ideally are very difficult to sort of access or do with other tech, right? That's the best place to intersect.
That might cause our list to look a little different than the prevalent list if you just did the literature prevalence from 1 to 20, but it will be targeted to what those customers are saying. I think more to come on that as we progress next year and roll out sort of the launch capabilities. Okay.
This is RK from H.C. Wainwright, I have three questions, so maybe I can go one at a time. We just talked about pricing and where it could fall. I'm thinking about the market. With the Platinum and Platinum Pro over the last two to three years, you have generated a market for this sort of product. How much of an overlap is there from the Platinum to Proteus in terms of the market itself? What additional market could you start tapping into with Proteus?
That's my first question.
Okay. I'd answer it in a couple of ways. I think if you look at those basic biology labs that have a Platinum or Platinum Pro, I think there will be some of those that will be of a scale to acquire an instrument in that price point, but some of those will obviously fall out. They're too small. They don't have that level of funding. I think some overlap there, but I think it's more modest. I think as we look at core labs, translational, biopharma customers, defense customers, when we look into those other segments, early sort of intel would suggest many of those folks, probably based more on the capabilities of Proteus than anything else, are likely to want to move to a Proteus platform. I think we're going to continue to have these discussions.
That's why we want to release certain information at certain times. I think the capabilities of Proteus is bound to be the main decision maker more than it is necessarily the segment or the type of work they do. If we just look at those pioneer grants, if we look at the types of things people are publishing on, it would suggest that there could be pretty significant overlap outside of the basic biology lab between that Proteus, that Platinum base, and the future Proteus base.
The next question is, Todd. He was talking about signal-to-noise ratio. If I go back and think about my days in lab, sample preparation is a huge part of that, I would think. How do you overcome that so that you can maintain uniform SNR?
Let me start, and then I'll give it to Todd.
The first thing I'll say is, on the highest value applications, I think our tilt at this point from our experience in the market is going to be towards having sample prep partners where we have in some way characterized and validated the performance of their reagents in front of our product. If you go out today and buy an enrichment kit for, name your favorite protein, and you buy it from three different people, I can assure you you're going to get three very different sort of purities, concentrations, what other things come along for the ride. We have an application development team that interfaces with customers daily, and we see this.
I think where possible and where it's a very important application or where we want maximum control over the performance in terms of the claims we can make, I think you'll see us have partners and steer customers to using certain kits in front of our technology. I think more fundamentally on the technology point, I'll let Todd talk about how he thinks about this aspect.
Sure. I think Jeff's answer is really the most important part of the answer because that sample prep aspect and sort of the noise that comes from other things and samples, that really does drive more into how we deal with the sample, how we get it into the system. The SNR that I presented in the presentation is about we've immobilized the peptides on the chip, and we have our binders in solution, and they're binding to the end-terminal amino acid.
It is that fundamental level of signal that we detect over the background noise in the system. That really has more to do with the background from the binders that we have in solution, the background from the imaging system. Those are fundamental aspects of the system we have a lot of control over. We really define that SNR component, and it gives us that very sense of detection. The other component from the sample prep is, well, we could have a bunch of off-target things on the chip that we'll also be sequencing, and we have to tease that out. A lot of that goes to how we prepare things that we sequence and the software on the back end, how we process it. I think they're a little bit two different SNRs.
Thank you both. The last question from me.
There is a lot of advancement which has been achieved, obviously, both on the recognizer side and the PTMs and whatnot. If we want to keep Platinum as the annuity going forward, how can all these improvements, can they be taken back to the Platinum and Platinum Pro and still be used in those missions?
I think there are two answers to that. One is from a sequencing chemistry perspective. A lot of the sequencing development we do is still on Platinum. We know these capabilities work there. I think the question we have to sort of work through as we lead up to the Platinum upgrade and sort of goes to your first question, which is, what's the overlap? Is really, what will that overlap be? And how many people are going to want to move anyways?
What then is the remaining base and what's the cost and value of maintaining now two different architectures, two different product lines? I think we don't have any definitive plans right now to say, "Hey, we're ending Platinum on someday." We don't have that level of thinking. I think we're going to let this sort of conversion or upgrade path probably inform that before we decide to make it backwards compatible in terms of doing the development to bring it on board. The fundamental chemistry, the fundamental algorithms, those things could work on either platform. One just looks at lifetime and one looks at color. They could go there. It's really more that question of the overlap in those bases. Obviously, there's a lot of benefits to customers to move to Proteus.
There's a lot of cost benefits for us to move people there. I think we'll weigh all those factors.
Thank you. Scott Henry with AGP. From a big picture, with a new technology, the challenge is always to find the inflection point where you could reach exponential growth, where things take off. It sounds like what I'm hearing today is that you think PTM could be the driver that has larger mass appeal than what the customers are asking for. It sounds like going from 2 million wells to 10 billion is incrementally helpful, the same with going to all the amino acids. Is that the right takeaway? Do you think that is the key driver?
I would intersect the amino acid coverage with PTMs because those things together sort of are amplifiers.
The better the coverage, the more data and sort of richness of information you get, the more things you can identify. If you take a step back, though, and maybe answer it from the customer perspective versus the technology perspective, I would tell you that PTMs are the thing that are of biological relevance and importance and are very difficult to do. When you talk with customers, yes, I can point you to a lab that has the highest-end mass spec machine, has amazing infrastructure, custom pipelines, all kinds of different things that can do a lot more than the average mass spec lab with the same piece of equipment. We do think PTMs is a major sort of opportunity in terms of its importance and it being underserved.
We do need to intersect the coverage with it to get sort of the bump we want in terms of performance. I think if we do what we described here today, that would be the jumping-off point we think for Proteus is really being able to do broad PTM profiling.
Great. Thank you for that color. Now, with regards to getting to 20 amino acids, what is the risk profile of that challenge? Is it blocking and tackling, or is it pretty predictable?
John, do you want to take? I'll give you my first answer, and then I'll let John add a little color, which is if we thought it was a big architecture leap like Proteus was, we did not talk about Proteus until we were comfortable that we understood sort of the risk matrix and the list.
While it was not de-risked when we came here last year, we had enough confidence that we could get there that we talked about it. I think similarly, is it zero risk? No. I think we were comfortable that we sort of have control over this enough and have sort of a robust process that we can get there. Maybe John can talk about how he thinks about it and manages through it with the team.
Yeah. I think what I was showing in the slides was we have demonstrations of how to get from A to B. We have done it. We can go get new binders. We can take existing ones, modify them. I think all pathways are open. I think I agree. The risk profile is low.
The only part is I can't exactly anticipate how it's going to turn out. I don't exactly know how it's going to go, but I'm confident that we'll get there. I've seen the internal data. We're already at 18. We have them. We have to just close the gap.
I think the hard thing for the guys and the gals doing this work to predict is always I think John laid out again sort of a, if you remember back to his presentation, he said, "Hey, here's the eight recognizers I think I'm going to need, but don't hold me to the exact combination." I think that's that sort of how question.
If we were not comfortable that we could do it, I think anyone who has interacted with the company for the last three years, if we did not think we could pull it off, we would not be sharing it here today. We would have taken a more muted position on path to 20.
Okay. Great. Final question. Just when we model out the launch curve for placements, it sounds like at some point, customers will be waiting for Proteus. Is that a fair assessment that 2026 Platinum sales will probably be flattish? There will be sales, but the bigger focus will be that 2026 launch. Any comments how you would expect Proteus to launch? Will we be early adopters, or will it see more wide scale? Just curious of your opinions. Thank you.
Sure.
Some of this, Scott, we'll bring out in more wholesome sort of updates as we get to some of these commercial milestones in Q3. I think the first part of your question, I think it's fair to assume that as more information gets out about Proteus, especially those external customer collaborators start talking about it, that obviously has a muting effect on the current platform. I think that's okay. We have a certain number of machines already built. We can use those via placements and get data generated and get consumable usage. Maybe we don't capture the top line of having sold an extra Platinum. If we have a user of the tech who then wants to upgrade, we'll get to capture that when we get to the Proteus launch.
I think it's a good trade-off to make for sort of short-term top line over the longer-term sort of potential to get more people onto Proteus earlier in its lifecycle. I think in terms of how to think about the modeling of that, we're not really able to provide guidance on that yet today. Again, I think as we go through these dialogues, as we start to release information, we'll get a better feel for who's going to be an upgrader versus who's a net new address. I think that'll play a lot into how we start to try to provide guidance to the street on what the adoption looks like. Because obviously, converting somebody where you've done that work, where they're already a believer in the tech, is a lower burden than it is to go get a brand new address.
I think we'll learn a lot over the next sort of couple of quarters and have visibility into what do we think that mix might look like at launch.
Got one behind you. Hi, Puneet Souda at Leerink Partners . Jeff, thanks for the presentation. I'll wrap my questions into one. When you look at the 20 amino acid detection, how much of a fidelity do you have across them in terms of performance? Are you detecting them at the same performance level? I think that's an important point. When you look at the PTMs, phosphotyrosine antibodies have been around for a while. What I'm trying to understand, are you combining kinetic with those to actually enhance the performance?
What does that mean for your gross margin, both in terms of what the output is and what it means for gross margin if you're going to employ more of those binders and antibodies? The last one around positioning of the product. On one side, you've got Orbitraps, the timsTOFs. You've got the low-plex amino acids on one hand, but the high-plex, you've got the PA and O-links and aptamers and other technologies. Just thinking about all of that, how are you positioning the platform?
Okay. Go back to coverage. I think you picked up on a very good, I think, sort of next-level detail, which is why we talked about demonstrating 20 and then having it in a kit.
Because there's a difference between sort of demonstrating and having enough of the coverage of that amino acid every time it shows up in the proteome. We think you got to have some cycle to convert that and get into the kit in 2027. I think John focused mostly today on just specifically how do we detect all 20. What he didn't talk about, but is something his team actively works on, is they're always looking to up the coverage of any given recognizer. Some of those recognizers have been evolved for a long time. They have very high sort of sequence context. Some of the ones have less. Those are actually being evolved in parallel. It's not exclusively just new. It's about just moving everything up to maximize all the sequence context you can see.
That also wraps in with what Todd talked about, which is sequencing depth. Those three things together maximize it when it's present, maximize the depth you can get, and detect all of them. Those three together is what gets you to sort of the maximum coverage. All three of those things are being worked on. We just sort of focused mostly today on helping people understand this was not going to take us three more years to detect all 20. We were going to get there sooner. In terms of the positioning of the product, I would agree with you that there are sort of different markets and different ways to cut it.
I would say our view on PTMs, on variants, and some of these other types of analysis in ultra-low abundant biomarker panel is really trying to focus our tech where we think it will have standalone value in the face of those other technologies. What we're not standing here today telling you is we're trying to figure out how to do 10,000 proteins from plasma to go compete with Orbitrap. Because we think it does a pretty nice job of that based on the data. We're not trying to necessarily today go compete with O-link who says, "I'm going to give you 5,000, 6,000 proteins from this sample and give you relative abundance." We think those are great. Those are needed. We think downstream of that, there's lots of work to now explore those panels they identify for PTMs and these other things that they can't do.
That is sort of the place to go fit in rather than chase a 5,000 biomarker panel that now we're sort of in the margin of, "Well, we're better at this. They're better at that." You don't really have that differentiation and some of the opportunity to maybe differentially price. That is why you hear us moving, again, both from our own assessment of the market, but also I think it shows up in the customer data. They're not asking us to give them a 5,000 protein panel. They're asking us to help them look at methylation, acetylation, and phosphorylation on this protein because they can't do that. I think competitively, it makes sense. What the customer is saying sort of ties to where the competitors tend to be really good right now. We are mindful of that when we're coming to market.
We'll let RK have another.
Actually, it's a corollary to the question that Scott asked. Do you need to have either 18 or 20 amino acids and enough number of PTMs for Proteus to be a useful technology, or what you have is good enough to get Proteus to be launched?
Yeah. I think the answer to that question is we're going to, first of all, the fundamental capability, and I'll tie it back to Punitza. I missed one part of your question and just sort of spurred my thinking now. The coverage component of amino acids, the better the coverage we have, the more of the PTMs we can do via kinetics. The less we would need a specialized recognizer or some sort of other affinity reagent.
When we use an affinity reagent, though, the nice thing is we need very little to be able to see it at a single molecule level. Some of the work Brian showed in phosphorylation, when we've used some of these off-the-shelf reagents, we can use them at very low concentration. It's sort of in the noise of the cost of goods of building a product. Our tilt as a company, right, our bias wherever possible will be to do as much of the PTMs through the kinetic signatures as possible. We use these other tools when we want to maybe optimize for a certain use case or capability the customer is asking for that we don't yet have in sequencing. To your point, though, RK, I think our view is the 20 amino acid coverage is the foundational thing we got to do.
That opens up how we apply it. When do we apply it towards PTM 1 versus 2? I think that goes back to my earlier point, which is we're spending that time with customers right now. Phosphorylation is obvious. That's why we show data on it. We've been doing a lot of work in the area. We've been doing work, as Brian showed, both in kinetic signatures and with pre-recognition. That all underlies what customers are trying to do, what some of the partners like Karna are trying to do. It is informed by that. Obviously, phosphorylation is something that's top of the list. How we allocate our time between now and the launch to bring on the other PTMs really comes back to this sort of prioritization we'll do with customers on which ones will add the most value to them.
We expect it to be more than just phosphorylation, but exactly what it is, we don't know that answer today. We are focused on what you're asking, which is what is the totality of capabilities? Because again, PTMs is a part of it, but if we can do an ultra-low like an interleukin or a cytokine, that's also an application very difficult to access with other tech. I think we're looking at the totality of things we can do, things we can partner to deliver to answer the exact question you're answering, which is make sure when we get to the Proteus launch, we have a wholesome set of capabilities to drive the adoption of the platform. We're definitely very cognizant of that, and that's where the focus is. All right. With that, I think that's the end of the questions.
Again, for everyone in the room and those online, thank you for joining us today. We appreciate your interest in Quantum-Si, and we look forward to future updates. Thank you.