Seer, Inc. (SEER) Earnings Call Transcript & Summary

All right. Good afternoon, ladies and gentlemen. Please take a seat. We're about to get started. I'm Julia Qin, Lead Analyst covering life science tools and diagnostics at JPMorgan, and it's my great pleasure to introduce you to our next company presentation by Seer. So with that, let me turn it over to Omid. Thank you.

Hello, everyone. Julia, thanks so much and to the JPMorgan team for invitation to be here. I'm Omid Farokhzad, CEO of Seer company. I founded in 2017 -- December of '17 and took public in December of '20. In January of '21, we brought in our flagship product, the Proteograph Product Suite in limited release to customers. And a year ago, this time, we brought the product into a broad commercial release. So today, I'm going to take you through some of the progress we've made over the past year, tell you about some of the commercial traction, what the customers are saying. And then share with you more about why I believe Seer is well positioned to open up a new gateway to the proteome and really open up the next frontier in biology. My presentation will follow a Q&A with David Horn and myself. David is our CFO. Note our safe harbor disclosure, which indicates that this presentation is going to have some forward-looking statements. So at Seer, our vision is to redefine what's possible by imagining and pioneering new ways to decode the secrets of the proteome to improve human health. Proteomic is the next frontier in biology, and we're focused on developing transformative products to really unpack the power of the proteome that until now has remained largely locked. This enormously powerful content is undiscovered and Seer aspires to change that and -- we're more excited about the opportunities that's in front of us. We believe Seer is in an excellent position to change the trajectory of proteomics and open up this frontier in biology and human health. To start, we have a technology that's broadly accessible and enable the scientific and the clinical community to access unbiased proteomic with a combination of depth, speed, breadth, scale that can potentially open up entire ecosystems. And I believe the potential of unbiased proteomic at scale may actually be larger than other omics today. Our Proteograph product suite is highly differentiated and it's also broadly accessible. And I say this because not only is it easy to use, it takes the black part out of the use of the unbiased proteomic. But it also fits within a researcher's budget. And the entire workflow sits upstream to a large installed base of a detector. This is a mass spec. Obviously, there's about 16,000 of them installed globally that does proteomic work, and that is growing at about an 8% CAGR. We believe the combination of these attributes will excel adoption of our system and importantly, not just by proteomic folks, but by the broader scientific community that is looking to add this molecular information to the types of studies that they're doing and the problems that they're looking to solve. Now there has been a tremendous -- a very large mismatch between our access to genomic and proteomic. And this was really one of the key drivers to start Seer. Access to this detailed and complex information is really key to understanding biology. Virtually every function in the body happens by an action of a single protein or a combination of proteins coming together to form a machinery and together to achieve that. With large-scale access to deep and unbiased genomic information over the last 15 years, we've now sequenced over 1 million genome, over 10 million exomes. Across the population, that's resulted in identifying over 1.1 billion genetic variance. Yet today, we don't know much at all a tiny fraction about the functional context of this information at the protein level, at the functional level. Now part of this reason is there's a gap in biology that is because biology is a dynamic complex, purposeful metrics of interaction across molecules, and that complexity goes even deeper than we had previously imagined. As you move from the left side of the slide towards the right, you go from 20,000 genes, there's a log increase in terms of the number of transcripts that you make all the difference in RNA processing splicing, you end up with 200,000-plus transcripts. And by the time we end up with proteins, your -- another log increase and there's millions of variants of proteins. Now, these are -- these variants of proteins originate from the same 20,000 genes, but they can have vastly different functions. Population scale proteomics is needed to decode the complex of the proteome and to really annotate the function of the genetic variants, meaningfully advancing our biological insight. Now there was a recent paper published in nature that looked at exome data from approx 450,000 individuals in the UK Biobank database, and the findings are really intriguing. They found a remarkable number of potential protein variants within each sample. The table summarizes their data. It underscores how little we know about the complexity of proteome. On a per individual basis, there were about 3,000 variants of proteins that are potentially deleterious. And in approximately 200 of those cases, those variants can cause complete loss of protein function. Now if you look at the entire 455,000 subjects that were sequenced, then the number of potentially deleterious protein variant is more than 6 million. Now this paper unequivocally underscores the unmet need to understand protein variant at peptide and amino acid level resolution. And if we do this, I believe that a massive impact will be made on diagnosis treatment monitoring of disease. Now plasma is the most accessible bio sample for population-scale studies. And prior to Seer, deep unbiased plasma proteomic at scale was impractical with the largest unbiased deep study comprising only 48 samples. And the deep study published by researchers at the Broad Institute was 5,300 proteins. Now Seer entered limited release in the Proteograph Product Suite in December of '20, brought commercial release a year later in January of '22. In this short time, we've seen customers scale their studies by orders of magnitude with multiple studies of over 1,000 samples completed to date, achieving unprecedented depth of coverage, something that was fundamentally completely impractical just 2 years ago. We're at a watershed moment in proteomics, the likes of which we saw in mid-2000s in genomic, where access to novel content progressively became possible at larger and larger scale and new markets were created or end markets were expanded. Now the dynamic range and complexity of proteins and biological sample necessities commercial workflows that take a lot of equipment, manual labor expertise and time that are not readily accessible to most labs and most scientists. These conventional approaches are fundamentally limited in their scalability and hence, why prior to Seer the largest deep unbiased proteomic was 48 samples. Seer unequivocally solve this problem. Our technology removes the complexity and enables access to proteomic content at scale, speed, depth, breadth previously not possible. Seer uses proprietary engineered nanoparticles, bringing together key attributes an unbiased approach, deep proteomic interrogation, a rapid automated protocol and scale that was previously not possible. Now the dynamic range issues, solved. As a result of these 4 attributes working together, our customers are now able to obtain highly accurate, reproducible, quantitative measurement at proteome across the entire dynamic range. Importantly, these measurements are always reported with a 1% false discovery rate. This is an important point. When we report, we see something, 99% of the time, what we report to see is accurate. By the way, as we change the false discovery rate to a higher number, the number of protein IDs goes up. In addition to our technology, it's applicable to a wide range of sample types and we'll work with any species, including model organisms typically used in medical research and drug development. So I believe Seer is enabling unprecedented proteomic access and novel biological insight. Now we're well positioned to become the definitive tool leaders in proteomic. Over the past year, we've more than doubled our installed base in the Proteograph Product Suite in the U.S., Europe and Asia. We've enabled our customers to drive unique insights with a range of biological research and translational applications kind of diseases such as cancer, diabetes, cognitive impairment, while empowering them to do studies at scale previously not possible. Seer's market opportunity is large. It's growing. The Proteograph Product Suite can be used to accelerate our understanding of biology and human health across both the proteomics and the genomic market driving demand and expansion across both. We envision a future in which the entire ecosystem and end markets could be created or expanded with customer using the Proteograph Product Suite across doing deep unbiased, rapid scalable proteomics for a myriad of applications. Now as I described earlier, protein variants play a critical role in biological function. They also impact protein structure and surface confirmation. This inherent biology creates a key limitation for affinity-based or targeted proteomic methods that rely on affinity reagents for detection. Affinity-based approaches involve a ligand, such as an antibody or an aptamer that binds to an epitope on a specific protein. A typical epitope is about 5 to 8 amino acid long. And a typical human protein is 472 amino acids long. On the left side of this slide, 2 examples of antibodies binds to 3 variants that the same protein are shown. The antibody that binds to the left side of these 3 protein variants binds to an epitope that is conserved among the 3 different protein variants and does not distinguish the 3 from each other. Conversely, the antibody that binds to the right side of these 3 protein variants may have this binding epitope disrupted. Now in the top, the protein variant where the elB is intact, the antibody relatively binds normally. In the middle protein variant, a post-transitional modification or an amino acid substitution, schematically represented there with that red circle, alters the confirmation of the epitope and disrupts the antibody binding. And in the bottom example, that protein variant, it's missing an entire domain from RNA splicing sample, and the epitope would not be present and the antibody cannot bind at all. In principle, targeted approaches are unable to distinguish protein variants and may inadvertently result in false conclusions. Now the middle panel in this slide, Pietzner, E.T. al from the University of Cambridge published a paper about a year ago in Nature Communications and experimentally demonstrated for you what I schematically shown on the left, the challenges to affinity-based approaches. This is shown in their figure, the figure under -- in the middle is from their paper. On average, the correlation between 2 commercially available affinity-based method was 0.38. By the way, the shape of correlation is bimodal. You can have complete correlation or a total lack of correlation on the average of about 0.38. And it's largely because of variants of the same protein interact differently with the different affinity-based methods. The paper really underscores the importance of looking at proteins at the amino acid and peptide level resolution with a technology that is quantitatively robust to protein variants. Now multiple variants that the same protein can arise from a single gene, as I said this through RNA processing, and these are called protein splice forms, which you can think of really as protein variants. This is schematically shown there for you in splice variant 1 and splice variant 2 in the slide. The majority of human genes produce more than 1 protein splice form. In fact, it's estimated that about 70,000 splice form are created from our 20,000 genes. Now that number is grossly underestimated because at a population level, a much larger number of protein isoforms exist because of genetic variants that alter RNA processing. So in short, the biological complexity is huge. Now affinity-based approaches are inherently unable to recognize the different splice form from each other, shown there on the upper part of that on the slide. Unbiased approach on the other hand, survey proteins at the peptide level or at the amino acid level and can distinguish different splice form from each other. Now to illustrate why this is critically important, let's look at a cohort of cancer and healthy subjects where you have 2 different splice form. It's possible that if you looked at the protein at the whole protein level using these affinity-based approaches, you would actually miss that those 2 different splice forms are different between healthy and cancer. On the other hand, if you had the resolution to see their splice form from each other, you may notice that one of them is upregulated in cancer but the other splice form is downregulated in cancer. Now I showed this to you schematically here, but let me share with you actual data. Now the important point is illustrated here from data from our non-cell lung cancer study that we published in Nature Communications. Here, we show 4 examples of protein. Each have at least 2 isoforms. And in each of these 4 cases, one of the isoforms is more abundant in cancer and the other more abundant in healthy, providing potentially important biomarker information. If you had looked at this information, these examples, at the whole protein level, you would not have been able to see the abundance of each protein variant separately missing this absolutely important biological insight. Now as I showed in the previous slide, unbiased approaches at the peptide level allow you to uncover protein variants and discover potential biomarkers. Each time you run a sample, you generate tens of thousands of data points at the peptide level, rare protein variants, post-transitional modification, protein-protein reactions can all be interrogated. The more samples you run, the more new content you discover and the amount of insight that can be gained is enormous, analogous to that 455,000 study published in nature from UKBB, where information in one subject showed only 3,000 protein variant, but across the 455,000, you saw 6 million of them. Now until recently, proteomic and genomic have been largely distinct fields that rarely intersected in recent years, though, the field of proteogenomics has emerged with the desire to bring proteomic and genomic information together in large cohort studies really enabling conducting genotype to phenotype. The Proteograph Product Suite is uniquely well positioned to bridge the gap between proteomics and genomics, accelerate the impact of proteogenomic and really contribute to connecting genotype to phenotype. Now one of the key goals of proteogenomics is to identify genomic variants that control protein abundance level. These variants are called protein quantitative trait loci or pQTL because the protein abundance level is seen as a "quantitative trade." These studies are powerful in understanding disease mechanism and finding suitable targets for drug discovery. To conduct a pQTL analysis, genetic variants are identified and then protein abundance levels are measured to identify those proteins in abundance correlates with the presence or absence of a genetic variant. When affinity-based approaches are used for pQTL analysis, these ligands bind to specific epitope of the protein shown schematically on the left side. Now I assume that the gene is being transcribed and translated to a protein and then that antibody would bind to that epitope from that exome that is schematically shown there. However, as shown in the middle panel, variants of the proteins can alter ligand binding because you can also have a genetic variant in the same exome where that protein would bind. Now this alter binding is often falsely interpreted as a pQTL and can be huge unrecognized problem in a study. With Seer technology, large-scale proteogenomics studies can be undertaking at scale in a highly accurate way. This is schematically shown on the right side of this slide. The variant peptides in red and non-variant peptides in blue. Because we quantify protein abundance at the peptide level resolution, we can accurately identify pQTLs allowing a better understanding of disease mechanisms and more successful drug discovery efforts when this is done at the protein variant level. Now affinity-based approaches essentially are required to have a tool, a hook, a ligand for each protein. But when you look at the complexity of the proteome, this is completely impractical. Unbiased approaches, grab a basket of proteins in an unbiased way, where you look at the totality of content accurately, quantitatively reproducibly and at scale of what is available in the biological system. Now looking back over the last year, I'm extremely proud of the Seer team that did exceptional work and the incredible work that they accomplished together. It was our first year of broad commercial release. And it's great to see the Proteograph Product Suite performing exceptionally well in the hands of our customers. The quality of the data being produced is fantastic. We have demonstrated the power and versatility of our platform and its use across different sample types and model organisms. We launched our Proteograph Analysis Suite 2.0 to enable proteogenomic analysis. Commercially, we more than doubled our revenue and installed base year-over-year and placed our next product in the hands of early access customers late last year. We remain extremely well capitalized with approximately $440 million of cash on the balance sheet, no debt, which can fund our growth for years to come. Finally, the most important is we continue to attract world-class talent and drive continued growth business. As we have delivered on these milestones, we've seen our customers expand the use of a Proteograph. We now have customers spanning academic research, translational, commercial, including pharma CROs, even applied markets. We're seeing the pursuit of a broad range of applications from catalog and protein variants to proteogenomics to multi-omics disease detection, biomarker discovery and clinical studies. We're excited to see what kind of studies they do next. Another key benefit of our platform is that the technology is inherently extensible in that it is species agnostic. And is able to analyze not only plasma and serum, but also other biofluids across human and model organisms. We measure the performance of the Proteograph Product Suite workflow across model organisms, urine, cerebrospinal fluid, conditioned media and compared to performance to neat biological samples. We observed gains in protein coverage across all different sample types. Importantly, in each sample, we measure tens of thousands of data points at the peptide level, providing information on thousands of proteins. We have continued to push on the next set of innovation on our software as well, expanding the capabilities at the Proteograph Product Suite. In Q3 of last year, in August, we released the Proteograph Analysis Suite 2.0 or PAS 2.0, a first of its kind proteogenomic workflow that maps peptide level data to genomic data to identify personalized variant peptides, not captured in canonical reference databases. Customers can more easily connect genomic data to proteomic data and assess peptide level disease association. Moving forward, we plan to continue to expand our software feature set streamlined data management across our workflow and lay the road map for larger scale population proteogenomic studies. With the Proteograph Analysis Suite, we've increased the computational capability for our unbiased proteomic analysis of large cohort by 1 or 2 orders of magnitude. [ Seam binding ] data analysis is a focus as we pave the way to more and more labs to adopt unbiased deep proteomic at scale. The feedback that we've received from our customers has been tremendously positive. In 2023 -- in 2022, I should say, we published 2 seminal papers characterizing our proprietary engineered nanoparticle technology and demonstrating how our Proteograph Product Suite flow has superior performance in terms of precision, depth throughput compared to conventional workflows. In our seminal publications, we have seen 119 posters at conferences, 19 of which are from customers, 20 oral presentation. We received recognition from the Human Proteome Organization, or HUPO, for science and technology innovation, and we were ranked #4 on Top 10 Innovation for 2020 by The Scientist. I'm also very excited about the traction we're gaining in the scientific community around our technology. Our customers and collaborators are able to redefine what's possible by leveraging unbiased approach. Josh (sic) [ Jon ] Brudvig at Sanford has been able to discover novel biomarkers for Batten Disease using pig as a model organism. Josh Coon at University of Wisconsin is onboarding the Proteograph Product Suite to redefine his mass spec methods for translational research. Neil Kelleher is collaborating with us on a new method to enable cataloging of human proteoform. And Jennifer Van Eyk has used the Proteograph to quantify clinically relevant markers in diabetes recently presenting a work at the NIDDK, TAMADOR 2022 meeting. There's much, much more to come. I have visibility to customer manuscripts being submitted for peer review and some of the data that we have seen is just tremendously reassuring and gratifying as a scientist to see. As you may remember a year ago this time at the JPMorgan Healthcare Conference, we announced the launch of the Proteogenomics Consortium in partnership with Discovery Life Sciences and SCIEX. And I'm thrilled to announce that Discovery is now up and running, having set up a new facility in Boston with multiple Proteographs and mass specs that support the Proteogenomic Consortium. They've announced last month that they're ready to receive customer samples. We look forward to supporting them as they scale their business in 2023. Now yesterday, our customer, PrognomIQ, which is a private multi-omic liquid biopsy company in the Bay Area that was spun out of Seer 2.5 years ago right before IPO, reported that they have launched 15,000 perspective clinical study of multi-omic program, core to which is deep unbiased proteomic. And of course, this is for detection of early-stage lung cancer and follows the completion of their current study, which was a 1,031 subject, largest deep multi-omic study undertaken to date by any organization. That is deep unbiased proteomics together with metabolomic, lipidomic, fragmentomics, copy number variants, methylation and transcriptome. They will share their data themselves, but let me just give you a brief look at the level of differences that they see between healthy and cancer across the different omics. They're expecting to present their data at the upcoming AACR and the ASCO conference. We're very, very excited about the publication of this paper and the further presentation and details of the data. I'm very optimistic that they may have a leading program for early stage detection of lung cancer. Looking forward to 2023, we will continue to drive execution against our core strategies, enabling breakthrough discoveries with the Proteograph Product Suite, demonstrating its power, catalyzing new applications and markets and continuing to build an industry-leading team. We are very much at the onset of our journey at the very beginning. And while much work remains, we're excited and inspired by the opportunity that lays in front of us. I'm incredibly proud of our team for the progress we made in such a short amount of time. It's been 5 short years since we started this company. I'm humbled to lead this organization, this amazing team. I'm inspired by their passion, their hard work, their dedication, that's allowed us to commercialize such a transformative product. In summary, I believe we have a technology team, the strategy to bring the next phase in omics to labs all around the globe. Thank you, and I'll turn it back to Julia.

Great. Thank you, Omid, for the great overview. Let's welcome David on stage to join the Q&A. So I can get us started. But if the audience, if you have a question, feel free to raise your hand. So you showed a lot of great examples of some of the early discoveries that's enabled by the Proteograph platform. So maybe just starting from the big picture, are any of these discoveries, where do they stand in terms of being progressed to the next step, right? And over what kind of time frame can we see that unbiased or high-plex proteomics really moving on to the clinical applications?

Look, I think the perfect poster child is looking at an organization like PrognomIQ. Now PrognomIQ was spun out of Seer. Seer Founder, Philip Ma, who obviously knew the platform well. He's the CEO of PrognomIQ. So in his case, there was not a need to test and become a believer. He understood the platform well. So he hit the ground running from day 1. Another large-scale organization pursuing multi-omics, core to which is deep unbiased. Very quickly, they completed multiple studies at scale. And the study of 1,030 samples is a good example of that. And they're now starting at 15,000 deep unbiased proteomic -- multi-omic study as well. Now if you then look at other organizations, other than PrognomIQ and say what is the slope or the velocity of them scaling up. A typical customer takes about 9 months to go from where the Proteograph comes into the lab until they're ready to kind of do any studies of any scale. We typically do a small-scale study in the tens of samples and then the data for that then supports doing studies that are multiples of hundreds of samples. We've had now multiple customers studies of 1,000-plus sample. PrognomIQ is the first that has now announced this study of 15,000 samples. But I suspect, for example, Proteogenomic Consortium that is a service provider will likely this year will do in aggregate studies that will far exceed what PrognomIQ has done.

As we think about your customers continue to scale up their studies to those 1,000 samples or even greater studies, are there any bottlenecks in terms of workflow, cost, informatics or...

Data analysis was a challenge. If you look at the workflow of unbiased proteomic, the bottleneck was always upstream workflow. Seer unequivocally crushed through that. Then the next set of bottleneck is then a massive amount of data that you generate and how do you process all of that. And PAS 2.0 took a big step forward in addressing that. Mass specs themselves are getting faster and faster, together with continued innovation on the Proteograph as well. We should be able to do studies in the tens of thousands or potentially even hundreds of thousands, unbiased deep proteomic leveraging platforms like the Proteograph.

Right. As we think about Proteograph adoption, you mentioned some upcoming publications that should really serve as external validation for the platform. Curious what kind of performance metrics will these publications validate? And can you just talk about how meaningful impact those publications will generate?

We -- I have visibility to some of these work. One particular customer shared data with us at our recent scientific advisory meeting that we had with some KOLs. This is a customer that would be considered a thought leader in the proteomic space, a user of the Proteograph. Previously, that individual had published the deepest unbiased paper in a paper in Nature protocol, that was a small study of about 16 samples. That paper got -- that investigator to a depth of about 5,300 proteins across his 16-sample study or 4,300 in any one subject. He reported to us that with the most recent study that he ran, a 300-sample study that he ran with the Proteograph, he's reaching a depth of 6,200 proteins. So -- and in his words, this is unprecedented, both in terms of the scale that he's able to do and the depth of protein coverage that he's able to do. So that, to me, is an extreme validation of the scalability of the platform, both in terms of depth, speed, size of the studies. We also have investigators that are looking at precision and accuracy and quantitative nature of what the Proteograph produces. And again, Proteograph puts out data with the level of precision accuracy that in the studies as small as 200 samples that are powered adequately to see a 50% change in a biomarker concentration and a typical biological change in a biomarker is actually much more than that. So this study is -- can be used for clinical utility, and we're going to see publications like that emerge from customers as well.

Great. Now shifting to some more near-term dynamics. You noted on your recent earnings calls that, obviously, in light of the macro, there's some prolonged decision-making or sales cycle with your customers. Can you maybe give us more color on what the current sales cycle like? How does it compare to historical? And then in light of that, are you having any customer conversations regarding maybe alternative models instead of capital placements or alternative pricing arrangements?

Do you want to take that?

Sure. So in terms of the macro and the elongated sales cycle, I think what we're seeing is an interesting dynamic. Certainly, the macro headwinds have caused people to just be a little more cautious. There's a bit of a barbell approach here. I think for -- pharma is certainly still spending, but for proven technologies and things, I think for kind of newer technologies such as Seer's, there is, again, a little more willingness to want to kind of test and see the data, right, see that third-party validation that you've been talking about Julia. And so we end up running what we call proof of principle studies, so our POP studies. And so they will send us tens of samples, 30, 40 samples for a fee, and we'll run the POP study and then we present them with that data and then they are able to use that and understand the performance of the Proteograph. So just adds a little time, right, to the dynamic. And I think if you layer on the macro to that, that just creates some more just conservative approach, I think folks have had, which we've seen. So again, with the publication of these papers this year and other presentations and just a continued growing body of evidence around the Proteograph, I think that will help with that. But that's certainly some things we've seen in the near term. And the second part of your question around...

Any alternative like pricing discounts?

Oh, pricing discount, yes. So really, what we've tried to do is take down the accessibility barrier. That's really why we again announced a year ago around the Proteogenomics Consortium. We do have some other centers of excellence that provide that it as a service model. So folks that won't necessarily want to bring it in-house, but do want to access the technology, and so we've seen that. We've actually done some service projects internally at Seer. We don't want to be a service business, but for some strategic projects and strategic customers, we will do a service project every now and then to do that to help them understand what the platform can do and the power of that. So it's really just trying to knock down the barriers around some of those accessibility issues in the near term.

Got you. More as a near-term bridge or lead generation, not a long term kind of...