How ConcertAI Came to Lead in Cancer Data
The Harry Glorikian Show
For January 30, 2024 episode
Harry Glorikian: Hello. Welcome to The Harry Glorikian Show, where we dive into the tech-driven future of healthcare.
We’ve now made more than 130 episodes of the show, stretching all the way back to 2018.
And I’ve got to tell you, when I look back at all the health tech and drug development companies I’ve hosted on the podcast, there’s an interesting pattern that starts to emerge.
After appearing on the podcast, a very large number of those companies have gone on to enormous growth and success in their markets.
Now, who knows what’s causing that.
I’d love to say that being on the podcast is like a catapult to success.
But no doubt, what’s going on is that we’re pretty good at finding companies that are already on a promising trajectory, and bringing in their founders and their CEOs for deep conversations.
Either way, I can’t think of a better example of this than Concert AI.
I first invited the company’s CEO, Jeff Elton, to come on the show back in July of 2021.
At that time, the company was already one of the leaders in gathering and analyzing broad collections of cancer data about patients involved in clinical trials for new treatments.
The company’s specialty is going beyond the very specific endpoints measured in clinical trials and looking to electronic medical records, genome sequencing data, insurance claims data, and other sources in order to build a more comprehensive picture of cancer patients and their journeys through the healthcare system.
That kind of cancer data can be very useful to companies trying to track the performance of their drugs after they’ve reached the market, and to researchers planning new clinical trials.
And in the nearly three years since I talked with Jeff, the company has grown by leaps and bounds.
It’s taken over management of more data sources, including the massive CancerLinq database formerly maintained by the American Society of Clinical Oncologyr.
It’s struck up partnerships with some of the leading technology startups, research centers, and drug companies working to beat cancer.
And it’s leaning hard into the new wave of deep-learning AI tools and their potential to help find patterns in vast amounts of cancer data about patients.
It’s probably safe to say that ConcertAI has gathered up more cancer data about patients than any other company on the planet.
And investors have been rushing to pour money into the company, on the conviction that data is going to be the key to getting more and better cancer drugs to market.
That’s certainly Jeff’s conviction too, as you’ll hear in the interview.
Harry Glorikian: Jeff, welcome back to the show.
Jeff Elton: Well, Harry, it’s a pleasure to be here. And I appreciate being here twice.
Harry Glorikian: Well, it’s funny because I’m like, I was just, you know, my brain was saying welcome to the show. And then I’m like, no, no, back to the show. Because I think the last time we talked was July of 2021, which was. Uh, almost two years ago, but it feels like. So much, you know, history has happened in that time with you guys. I mean, I want to spend a few minutes so people who listen to the show regularly, uh, have an idea of everything that’s changed. So I think, first of all, there’s been very big step changes at ConcertAI that seem to indicate that you’re one of those rare venture backed companies that’s actually hitting the ball out of the park. I mean, you raised $150 million in new venture capital last year on a valuation of $1.9 billion. I’m almost sorry I don’t have stock, but, um, you raised $600 million overall. And I read that you had $160 million in revenue last year. And I know we’re going to talk about the CancerLinq acquisition in a minute or so, so let’s leave that aside. But in broad terms, how would you describe everything that’s changed at Concert AI since mid-‘21?
Jeff Elton: Yeah. So I think, um. You know, sometimes… What changed? I can also sort of say what was the kind of backbone and kind of provided consistency. And then that also kind of helped direct a little bit about what was changing. So, you know, what’s consistent is, you know, our name has had AI in it from the beginning. Right? So we were born in an AI company back even in late 17, going into 2018, etc., and so even started delivering our first solutions in 2018 that AI. So we’re like, you know, the grandparents of the A.I. industry, if you will. We had a concept that was true then that’s even more true now that you can’t be an AI company if you don’t actually have access to data. And if you’re going to be an AI company with data, you need to make sure it is large volume access to data, because you want to make sure that what you’re training and building on is generalizable to the classes of problems you’re going to apply the AI to. Otherwise you just, you know, you build in bias and you build in the biases can be negative biases. Or they could but they’re biases. And you know, they may actually kind of lower the utility of what you’re doing. So you know, we always had those kind of assumptions at the beginning. We were born a health care focused company. From the beginning, health care for us meant from the beginning, also sitting in two ecosystems, health care providers, research entities, and life sciences company biopharma innovators.
Jeff Elton: So that that was kind of consistent. So when you’re early and you’ve got some principles, you tend to be very, looking. And I know, Harry, you kind of built things through your life and career. You’re looking for those economic opportunities. And the economic opportunities become a validation of your strategy. And it’s, you’ve got a very high customer, almost kind of deal, project-based kind of approach. As you mature and as you begin working with multiple customers, you almost have to kind of force yourself to sit back with a discipline and kind of say, are the things you’re doing really sitting in individual markets, and can each individual solution and activity represent $25 to $100 million worth of profitable, appropriate margin revenue? Otherwise it’s not sustainable and I mean not sustainable as a responsibility to your customers, because if you can’t sustain it, then they can’t actually build off the back of it. And so we started making decisions almost around the time that I talked to you at that particular point in time, that what we do has to be things that we can do at scale. And when we do it at scale, it has to be something that can move into a core part of the business processes of our customers and our partners. So we had a lot of discipline of things we would kind of discontinue. We had a lot of discipline about the things we kind of built around. And we also kind of told ourselves, you know, you got to validate the market.
Jeff Elton: So we actually are kind of proud. We have an incubation, formalized methodology and culture that forces vetting of those early stage innovations. And we’re always enamored by our own innovations. But even the, you know, enamorment shouldn’t necessarily lead to sustaining it as something that you build a business around. And so that ethos has kind of kind of guided us a lot. And I think we force ourselves to vet things quickly. And then if things actually look like they deliver high utility, and we’re working with leaders in the field to confirm that utility, we try to kind of improve it, advance it, scale it, and kind of move it forward. So today we have four lines of business. They’ve kind of gelled and matured now into some very, you know, well defined areas. You know, you would imagine just from the way I’ve said it, all of them have to kind of, you know, show that they have the value to be $25, $50, on their way to $100 million each. And that’s true. Each of them now, we believe, has those kinds of characteristics for doing it.
Jeff Elton: One of them is our advanced real world cancer data solutions in oncology, hematology, urological cancers. We have what we believe to be some of the most powerful, research capable, definitive cancer data solutions. They’re scaling very nicely. In fact, as we go into this year’s J.P. Morgan, you’ll see the next round of some of our announcement of who uses. We’ve been through two renewal cycles, and I’d say we’re at about an 80, 90% renewal rate on everything. So obviously, you know, we like to think that gives people utility. But we’re very, you know, centric about evolving them.
Jeff Elton: Our digital clinical trial solution is our second area that has, we’re the opposite of a decentralized trial solution. We’re a highly centralized trial solution. So we’re we’re deeply embedded into the workflows of research sites. Very oncology hematology. You know, we are kind of moving towards high automation, identification of patients for study eligibility. And then we’re automating how you run a study. We pre-populate all the forms, and we’re actually trying to get ourselves to a zero data entry model for research teams running clinical trials. So it’s all extracted out of local cancer data sources. And that has a lot of benefits. It lowers the burden on the team, allows them to spend time with the patient. It allows them to run more studies on the same research resource base. And these people are hard to find and hire and do things. And the study, the cancer data itself, the study package has full source data provenance. It does not have the errors and the other things in it. So it’s a higher veracity cancer data source.
Jeff Elton: Our third area that we do work in is medical imaging. We actually have one of the more broadly deployed 1900 provider sites where our Terarecon imaging solutions deployed for doing primary interpretation. More than 50% of what we do is cancer related interpretations. And now we have one of the most widely deployed clinical AI layer that’s an FDA approved class one on the way to being class two medical device that actually deploys software as a medical device and 510 K solutions, kind of, you know, if I take you back to your diagnostic days and things of that nature, it’s kind of it’s now it’s kind of like it’s software devices for doing diagnostic activity and looking for features in that patient population. We’re creating a research version of that solution for developing AI models, a research version for finding features in patients with impacts. That research solution will now also run digital pathology. So for us these are just different images to apply AI to and kind of kind of bringing that together.
Jeff Elton: And then our last part, the fourth area is our we do a lot of work on patient identification and ensuring they are onboard to treatment and predictive adherence monitoring. So it’s kind of what I’ll call a treatment assurance kind of activity. So when you think about that we’re again right between the biopharma innovators and kind of those providers. And part of our job is to kind of be that middle, be that neutral. It’s we’ve you know, that’s why we have no conflicts to disclose. We’ve never taken in corporate money to kind of run it. We’ve never done an exclusive deal with anybody. We’ve tried to maintain that kind of health care neutrality. And we work with people that sometimes are competitors and, you know, so it’s just it’s something you have to do as a health care, health care technology company.
Harry Glorikian: So in the middle of all of this since the last time we talked, right. Uh, generative AI comes on the scene November, December of 22 with the release of ChatGPT by OpenAI. How is that, um, or how is all that progress or that, you know, attention to large language models or other foundation models changed, enhanced, impeded, the environment you’re working in.
Jeff Elton: So, you know, fortunately, you know, generative is a class has kind of been evolving. I think GPT 3.5 and 4, obviously in 3.5, kind of started providing just a completely different performance level and kind of set of capabilities and started going into, you know, being more available to end consumers so people could experience it. So one is we do believe that generative class solutions and there’s, you know, there’s multiple of these, right? Because, you know, Meta has theirs and there’s different versions that kind of come through. These are pretty these are powerful tools that actually have extraordinarily fast processing capacity. So it’s not just what, you know, what’s actually it’s kind of why the Nvidia’s of the world are here too. Because processing capacity combined with cloud infrastructure, combined with cancer data accessibility, combined with the maturation of these classes of technologies, have kind of beneficially converged here a little bit. So it will change how all solutions work, no doubt about it. In fact, you know, we’re going through a process and actually evolving re-architecting a lot of what we do. And we use AI, generative and large language models, for our internal production. So it actually runs parts of our company in addition to being in our solutions. So for us it’s, you know, it’s part of how we operate. And it’s also part of what we deliver in terms of what’s there.
Jeff Elton: So, you know, we believe some of our solutions will probably pivot to being a generative interface. So you no longer drop and drag or, you know, even if it used to be a nice visual interface where you didn’t have to type anything, we’d, you know, kind of get you a lot of places, but we’ll end up kind of moving towards a generative prop and interactive window and even things that actually have generative authoring of the answers and the results in a way that gives utility to some of the workflows that people are working on that part, we’re already kind of evolving and advancing quite a bit generative itself. And you’ve probably heard a lot about this notion of, uh, hallucinations. So models self-train. So these are self-learning, self-training things. You know, AI doesn’t really like empty cells. And so the imputation process, which is another word for the hallucination, is something that occurs because it actually needs to kind of infer something in a place where something’s not. So you can get around that in a number of different ways because generative is phenomenal at contextualizing the information it’s seeing. And with the establishment of that context, you can bring large language models and other actually AI based approaches up to that. Uh, we’ve actually done generative, views meaning run three classes of generative different technologies in parallel on the same underlying phenomena.
Jeff Elton: And when you do that, you don’t actually have the same problem with any of the inferred or hallucination, and you can actually even begin to move to an even a higher accuracy level on certain things. So we’re going to, you know, a lot’s being learned. We have a whole team that does nothing but run generative experiments right now on different classes of problems and different solutions. And from our perspective right now, sometimes it actually changes. In fact, the challenge I’ve had to bring back to the team, it’s not do you apply generative to what I’m doing? But can I take what I want to do and rethink it as a generative-native solution? And if I do that, what does that look like? And it may not even be the same architectural principles to what we had before. So it’s not layering generative into legacy. It’s actually allowing the power of what’s there. But at the same time, you can imagine from the description I gave you, I have certain things which are non-regulatory query analytics. Well, there I can be pretty generous with my generative applications in terms of what’s there all the way over to. I’m in a set of workflows that are either pre regulatory or under some regulatory standards.
Jeff Elton: And so when we do that we’re having to develop our own kind of, you know, rules of conduct transparency. You know there’s now models that tell you what your how your model learned. So you almost can like reverse out, what is it that it actually decided that it was going to allocate and prioritize, even in the configuration of what models that resulted at the end? And some of that you’re going to have to actually do, even as part of your process of having conversations with regulators, much less users who insist on having some of that same kind of transparency. But at the end of the day, classic solutions, things we can do, it’s going to collapse time. It’s going to actually kind of take it’s going to take tasks that, you know, took hours, days, weeks and just almost collapse that down to nothing. And it’s, uh, you know, it’s a, I mean, one we’re happy to be where we are and just surrounded with data. I mean, we have more cancer data than anybody else in our domain by now, you know, 34X over our nearest largest one. And so from that perspective, we you know, we’re aware that this is an area we can operate in and be a better operator in. And so we’re very excited about its potential.
Harry Glorikian: Yeah, I mean, I’m using these tools every day. And the most basic thing I do is, you know, theoretically a little bit more advanced and. Yeah. Oh my God. I mean, yeah, it’s it is uh. It’s. And every time I find myself, I’m like, when am I going to stop saying wow? Right. That that’s the funny part because they keep they keep just getting better and you can do more things with them. But. You know, the last time we talked, I don’t want to forget. Like we were smack dab in the middle of the Covid pandemic, and still waiting for the FDA to approve the vaccines. Yeah. So I’m curious. How did the pandemic experience, you know, change, say, the state of cancer diagnostics or how did it change what, you know, you guys were doing at Concert? Because I have to believe it had some effect.
Jeff Elton: Yeah, so I will I’ll start with a caveat of, uh, we’ve never had a down year in our history, so we’ve, uh, always grown year over year and had pretty consistent performance improvements. And, um, you know, we’re we’re a digital company, right? So we’re, you know, we’re based on cancer data and we build, you know, data models, we process data and we built software solutions. And so as a company, we we’ve always been kind of 100% cloud. And you know born cloud and born kind of digital infrastructure and capabilities. And a very odd way. In the beginning I kind of was almost feel a little awkward to say it just because, you know, we saw so many other companies and things just had real challenges in the beginning. Um, we were able to pivot as a kind of, uh, remote, uh, very easily. Um, we all of our partners and providers, uh, signed the ability to allow people to process and do things they did before, which were in kind of close surveillance and very secure, uh, physical facilities to allow that to be done in remote facilities with additional kind of security and things being put on it. So we were quite fortunate that the entire ecosystem allowed that. Um, in the beginning, we did a lot of Covid related things. We did a Covid registry, and we’re doing work with ASCO. We started actually analyzing, uh, different outcomes of patient populations. We were seeing discontinuities. Patients weren’t getting access to clinical trials. So people saw us as a solution. Being able to do some things during an interim period of time that they weren’t able to do other things.
Jeff Elton: So to a certain degree, we actually saw broader adoption of some of what we were doing during that period, just because you couldn’t do things the traditional way. So again, you know, you sort of you never want to look at something like a pandemic that lost so many lives and kind of created such havoc on kind of, you know, different industry sectors and things of that nature and say, you know, maybe we were even somewhat a beneficiary of that. But in certain ways, the pivot forced a speed of change in ways that it changed that where we were somewhat advantaged, you know, kind of in doing that. And so that’s persisted the nice, the fortunate thing. It’s kind of we always were to use the word nice when you do anything with the pandemic. But um, the, the fortunate part was that some of the ways of operating did prove to be more effective and actually persisted. So some things did not carry beyond the pandemic, and some things did carry beyond the pandemic. And almost all cases, most of our major solutions and activities continued beyond the pandemic. And if anything, you know, we’re getting people to try to standardize on them a little bit more. So I would say, you know, born digital and born to scale. And we were able to do that during that period, and we were able to substitute from some legacy ways of operating that now actually can become a more efficient way of operating.
Harry Glorikian: So yeah. Yeah, yeah. I mean, sometimes you need a good something to shake up the system so that, you know, change can happen as opposed to people saying, this is the way we’ve always done it. Okay. The really big news recently, I should say, because there seems that there’s a continuous wave of big news is the acquisition of Cancerlinq from the American Society of Clinical Oncology. I mean. That seems like a move that could be transformative for you and for CancerLinq um, you know, um, but maybe you could help the listeners understand. Say, I don’t know what this means. Could you start off by describing what CancerLinq is, why it was created, and what services they offer in oncology, you know. Yeah. Just to give people a grounding. Yeah.
Jeff Elton: So, you know, start with the caveat that if you think of my, our description is we’ve always been very close to the provider community and also to biopharma innovators. Uh, CancerLinq was was an entity that we actually had a contractual relationship with that went into place, um, October or so of 2017. Um, at that time, it was a wholly owned operating entity of the American Society of Clinical Oncology, kind of I think it was kind of formally start incorporated about 2015, and some of the concepts for it actually went all the way back to 2012. Um, CancerLinq itself was originally born as a clinical quality, and today so, so there was a set of quality scores, um, and quality and outcome scores, performance scores that bore the initials Q, O, P, I oftentimes called colloquially Qopi, and Qopi was developed by Asco. And it was uh, oftentimes required manual scoring of cancer data and patient outcomes. And originally it was conceived of around chemotherapeutic administration. And if you know anything about Chemotherapeutics, they are toxic at certain levels. And, you know, they’re designed to and they’re called cytotoxins and they’re designed to go after rapidly growing cells and kill them. But the administration of them needs to be done very, very carefully to make sure that you’re only going going after cancer cells. And, and so just all sorts of parts of that. And so because that was the mainstay kind of for delivering that therapy uh, delivering cancer therapy, that was a real focus for, for Asco.
Jeff Elton: Well, when they found it was such a burden to do quality and outcome scoring, they built a subsidiary to do the automation. So CancerLinq was created as one of the really first non EMR specific kind of cancer data organizations for cancer care. And so like ten different electronic medical record manufacturers kind of areas, you know, where we’re kind of represented across all the different providers. And about 100 different providers signed up in that program. So you have the Epic, the Cerner, the onco EMRs, the Inomeds and all these different EMRs coming into these individual providers. And to do the automation, the different providers, uh, provide access to their cancer data so that the Qopi automation layer would run. And in that context, they also allowed research uses and other quality of care research to be done on some of the data that came into CancerLinq. Originally, ConcertAI and Tempus were contracted by CancerLinq to provide sets of services to CancerLinq, which also included data curation and a variety of other things, and then could actually conduct research off of some of the CancerLinq data. And I’m using the word research quite intentionally, because there are obviously a range of restrictions that, you know, anytime you’re doing work under a business associate agreement, if people don’t kind of know all that language, business associate agreement is a set terminology and a concept for handling protected health information. And the rules are established by the Health and Human Services Department of the US government and operate under. Everybody probably knows what HIPAA is because you sign off on it, but it allows you to kind of handle that information but in appropriate ways. And it indicates if you’re under a BA, you can only do research and quality care work. And so we actually knew the data very well. And we did a lot of research on that data. And it would never have pharma commercial purpose use at all and things of that nature. Um, it turns out that actually being in the data business is very expensive. And CancerLinq was not a was not trying to be a revenue positive or profitable entity, and it required ASCO to invest a lot more than the operating contributions that it got from ConcertAI and Tempus to kind of keep it going and honor its mission of quality and research. And it did do that for a long, long time up until, you know, kind of a week and a half ago or so and did do that for a long time, but it’s for a not for profit to kind of support something like that in an ongoing basis. Um, it’s just, uh, you, I mean, even just you talk about gen AI, the technology is being applied to cancer data and that drive this are changing so rapidly.
Jeff Elton: And one of the features even of AI and gen AI in particular is it’s not cheap, actually. It’s actually expensive cloud processing that actually is required for doing that. And if you’re going to keep up with the state of the art, and if you’re doing research and quality and all this other things, you have to keep up to the state of the art or you lag, you know what’s out there. So CancerLinq, Asco went through a very long, deliberate process, decided that it would spin the CancerLinq out as a separate entity. It’s put a couple requirements that it need to remain a distinct entity. But it uh, sought a competitive process to, uh, kind of put it into the hands of a for profit entity and run it on behalf of ASCO, but also, you know, other potential applications of that. Uh, so through that competitive process, we were quite pleased we were successful in that process. And, you know, we were at least down selected the final two and then finally became the final one and kind of doing that. So inside of this, we’ve uh, we are assuming responsibility for the CancerLinq entity. We’re upgrading all the cancer data integrations, we’re upgrading all the cloud environment. We’re bringing some of the latest AI.
Jeff Elton: Uh, we have a multi long term agreement with ASCO now. So we are providing ASCO an ongoing access to and in fact broader access and larger access to research grade data to power ASCO research. And so they’ll be using our Patient 360 solutions and any ASCO aligned investigator at no cost. So this is you know this is they have it on this is a part of our investment to the oncology community. So there was, so there is research at a scale that even ASCO has never had access to before. You know, we are the largest research data company in the world. And now with this you know we can provide ASCO probably the best data of any ASCO affiliated investigator. So that means all academic researchers that are sanctioned and thought of as being bona fide researchers in ASCO’s definition, not ours, are eligible to have access to the data. In addition to that, ASCO has generated their next generation of their quality framework. It’s now called ASCO certified. And so we’ve committed to a two year. And they’ve committed and we’ve mutually committed to an ongoing process of technically enabling that and automating that. It’s an important framework because it updates what’s in scope, it provides clinical pathways, but a lot of really important patient specific measures about the patient experience and receiving care, which is, you know, an appropriate and very important part.
Jeff Elton: But all these also lay the foundation for value based care. And there’s actually a lot. Now, the latest incarnation of value based care is coming through some of the federal policies and things that were came back in the Inflation Reduction Act and some things that HHS is going through, but also private payers are actually interacting on that. So the quality framework is the foundation to value based care. And value based care is how actually health care providers will be able to sustain themselves because of the kind of margin declines that have occurred from traditional reimbursement. So for ConcertAI one is, you know, clinical quality. One, it does now bring together all the cancer data we and things we were doing before. And it is it’s it’s continuous with our history. Meaning we’ve done work with CancerLinq in the past. So there’s a, it’s not so much discontinuous. But what it does do is by far this is the largest quality of care research kind of capable entity in the world by, by, by a fair distance. And it is probably the only absolutely neutral, pan setting of care, pan EMR data source that exists. I mean, because, you know, if you think about there’s other companies that have done great work all the way back to like, you know, Roche owns a company called Flatiron Oncology, which even Amy Abernathy and others were kind of worked on. But it is an EMR. It’s one EMR and it’s kind of owned by one company, and you have another one that’s owned by McKesson and sort of others. But this is completely neutral. And again, ConcertAI has no pharma funding in IT, diagnostic company funding in IT, no payer, no provider. And we’ve worked very hard to keep that neutrality and that around doing that. And so there’s an ongoing partnership with ASCO. But you also now have a research entity that can truly kind of provide the type of cancer data that can, you know, power the next generation of evidence. Um, we’re also now bringing TrialLinq, which was a concept that CancerLinq had, which was patient identification for clinical study eligibility. And it’s something that they, CancerLinkq members want to have access to. We’re using our digital trial solution that will start powering TrialLinq. Uh, a lot more of our technology for analytic tools will be available to the CancerLinq members that will power side by side with the next generation quality tools. So start thinking about it as this ecosystem of technologies and infrastructure that becomes part of the operating infrastructure for the next generation of cancer care in the United States. And, you know, potentially ex-US. ASCO itself is a very active international membership and things of that nature. So obviously we’re interested in that as well.
Harry Glorikian: So, Jeff, just so people get an idea. I mean, is there metrics that you guys have looked at like, you know. Number of patients that you had in your system, how many were in ASCO when, you know, when CancerLinq and then when you put them together, you get I mean I’m sure there’s some numbers somebodies oh yeah. Thrown together that you might be able to share.
Jeff Elton: Yeah. So there’s about about 8 million, uh, 8 million lives now that are cumulatively that are brought together that are both oncology, hematology, urological, cancer data, and probably the next largest cancer data source is probably 25% of that or something of that nature. So, you know, on that level, this is, you know, for X such typically would be there, uh, there’s 100, you know, there’s 100 sites. And to that 100 sites, uh, we will probably also have about another 100. So you’ve got a very large footprint. This includes small, medium, large sized community practice, academic centers, NCI designated centers and regional health systems. There is no other cancer data source that goes across all those different settings of care, uh, either at that particular level as well. So on all of those levels, it’s a pretty significant kind asset for the research communities. For ConcertAI, it’s a responsibility and it’s a tremendous opportunity, and it’s one that we’ve, you know, we went into kind of, uh, it does allow us to conduct and do research and do activities and advance some of our strategies in ways that, you know, we can kind of do quite securely. I think we couldn’t be more thrilled that the new CancerLinq entity has the long terme relationship with ASCO. ASCO is by far the most professionally managed and highest impact medical society, medical society in the world. And you know that’s tremendous. And for the CancerLinq members. And ASCO has been working very you know with us side by side. It actually allows them to get access to things, you know, much more seamlessly than they would have, you know, than they were able to before, which was a resource constraint. It wasn’t a will. It was just a resource constraint. So. So.
Harry Glorikian: Not that you guys have been sitting still, so CancerLinq is just one of, you know, these major alliances or expansion you’ve announced in the last couple of years, I mean, right. Uh, for example, I think you guys, uh, recently unveiled a partnership with Caris, right, in Texas, and. I don’t know when I was reading it, it seems. It goes beyond a typical research collaboration. I mean, it almost sounds like a merger, or at least in in the scope of the cancer data that you’re sharing. Yeah, a merger.
Jeff Elton: Of data. It’s definitely not where, we’re two distinct companies and there’s no overlapping equity or ownership. Um, and yeah, I mean.
Harry Glorikian: You know, I when I think about Caris, it’s like. You know, it’s a company that does whole exome sequencing, whole transcriptome sequencing to gather data for machine learning models. Right. Uh, which then they apply to the development of new precision medicine approaches to cancer. So. How would you describe Caris? I mean, what do they bring to the table? Yeah. Um, and what are you bringing to the table? How does this partnership really, like, move the needle? Yeah.
Jeff Elton: So, Harry, great description. And I you, uh, you’ve touched a lot of different diagnostic parts of the world in your past life, too. So Caris is, um, very fast growing diagnostic company. They’re very unique in what they do in the clinic. And they next generation sequencing area. And you highlighted it nicely, which is right now no one has their standard panel. Does a full exome 29,000 kind of genes. Usually we’re talking about panels of 700, 1200 you know maybe a couple thousand. So you have pretty much you’ve got 29,000 full transcriptome um, whole slide images, digital pathologies also available, which we also can then connect longitudinal clinical records. So if you kind of think about, you know, what we think of this is doing is, um, there’s a firm that’s starting to get used a little bit of almost called causal biology. And we are actually building intentionally multimodal data sets different modes of cancer data collected around a kind of a patient’s entire patient journey. Why would we want to do this? Well, multimodal data sets and causal biology says I can establish with confidence insights into a disease process. I can establish it’s not correlative and it’s not I’m not finding associations. I’m actually establishing it literally how a process works with confidence. Or if I’m seeing response or nonresponse of a patient to a particular treatment, I’m actually establishing causally why that patient responded or why they did not respond. Was it acquired resistance based on the prior drug that they were treated with? And kind of, you know, stage two, was it you know, what what was it that kind of brought that about? And when you take this data and this information, it’s changing how translational science is occurring, because now I can take cancer data and information that’s kind of coming from my in vivo in vitro work, but non-human work, and map it over to now that I’m able to look at things with the depth of what I can bring on the human data, and I can determine what strategy am I going to take. And if you take things like, you know, ADCs, which may be therapeutics that kind of hit two targets at the same time or different analytic approaches. What I’m trying to do is I’m trying to actually stop a pattern of resistance, while I’m also hitting the targeted tumor of interest to give that patient the highest likelihood of the longest duration of responses possible, which also improves the likelihood that it’s going to be a better drug by the time that they actually that there may be a recurrence because, you know, resistance will happen.
Jeff Elton: It’s just I mean, our, you know, human cells, whether they’re cancerous or non cancers, have built in redundancy. They will build up patterns, resistance to something because it’s unfortunately kind of you know, how the how the biology portion works even on the bad cells. So when we’re kind of going through this process you know we we really started taking a look at their capability, our capability, our AI, some of their AI. And we kind of thought our data and their data when it brings together, it kind of creates a unique asset at a scale that was never possible before, and a consistency with how the cancer data is brought together. That really begins to kind of change how translation science teams, early clinical designing those first in human studies, acquisition of data differently over the course of the studies all the way through to even, you know, kind of post-approval kind of different strategies. So, you know, part of the reason and in fact, I think what you’ll see in what we’ll be presenting as we kind of go into this year’s J.P. Morgan conference, is this idea of a clinical genomic biotechnology field. So we kind of think of ourselves as kind of evolving from, you know, you start off as a company that has cancer data that generates insights up to the fact where you can now create data sets, tools, technologies and infrastructure that changes how you work and it changes the likelihood of a program success.
Jeff Elton: We’re trying to actually be so partnered with these people that it changes what programs they take into the clinic and the success of those programs through the clinic, and it actually provides information confirmatory that helps aid giving better patient care and treatment and obviously improving those outcomes a lot. So it’s an integrated solution. It’s presented as an integrated solution. It’s actually I mean, very rarely do you see two companies together that actually create one thing that’s delivered as one thing to our customers and the experience that they’re telling us. And you will you’ll see some additional sort of you know, who some of the the partners are beginning to kind of use this, but the experience are telling us it it is the approach we’re taking that’s also part of its utility to them and actually helping them change how they operate. So, you know, it’s kind of like, you know, can we think like a biotech company too? Although, you know, we’re again, digital, right. So it’s all…
Harry Glorikian: Well, I think, you know, when you, you know, I’ve always thought that when you build these systems. Yeah. The information is of value to, from the clinician who wants some of this cutting edge information about how to manage and treat their patient, to the person that is now creating the next product and. You know, it’s just a question they’re asking of the data. Correct. So I think, you know, I’ve always thought that this bigger, integrated solution is what’s going to get us much farther, much faster, as opposed to silo silo silo silo. And I can’t translate it later on.
Jeff Elton: Absolutely. In fact, Harry, just to build on your own statement a little bit, because I 100% agree with the way you just expressed it. Um, take use the word signatures before. So signatures would be a model derived over both from the transcriptome with maybe, you know, a pointer over to a molecular specific condition, but I can pull a model off of the transcriptome and kind of RNA portion of activity here a bit in terms of looking at that, that signature can it will have significance to a treatment decision. So I think Caris has a wonderful example in colorectal, which is Folfox/Folfiri. Which order and what sequence do you give the two agents to which patients? And it matters, but they can predict which order you should give it in for the optimized outcome, uh, for that individual patient type. So if you think about that, this sort of example now where I’m bringing information as models that’s clinically actionable, that’s been prospectively validated. And now go back to my business, I have Terarecon where I’m doing AI models now operating on medical images, radiological images. Now for research purposes. We’re actually beginning to do that on whole slide images or digital pathology. Now you take a look at these signatures kind of coming off of transcriptome overlaid with exome, which allows us to take very narrow, very specific treatment decisions with prospective validation, provide a model and a tool for doing that.
Jeff Elton: So you start getting these decision augmentations that can be brought right into the workflow. I’ll put another thing out there. We’re doing a program which we’re kicking off in January. And, you know, maybe you’ll invite me back to be a third repeater here or something on your show, because at some point I’ll, I’ll be able to unblind some of this stuff, but I wanted it’s on this theme and we think it’s one of the coolest things we’ve been able to kind of, you know, we have to we I mean, we’ve we’ve shown it can work. We have to do it at production scale. But there’s a very rare cancer that is devastating for the patients that have this one molecular mutation. It is and their outcomes are, it’s you know, it’s it’s a solid tumor. But with the current treatment, patients end up kind of also presenting with later on brain metastases. And it’s a it’s a very poor response and non-responsive to current standard of care. There’s there’s new medicines coming through that hit that target that are actually based on breakthrough designation status of some of the programs really offers some real benefit.
Jeff Elton: But the problem is next generation sequencing you’re so late in the patient journey that by the time you get there, they’re relatively decompensated. And it’s their treatment options and ability to tolerate treatment low. So we found that that mutation is predictable off of whole slide images digital pathology pre NGS. So part of the program we’re doing is we’re actually now collecting all the whole slide images. No NGS done running the AI model, finding those that have that sub mutation at a level of accuracy and recall whatever threshold level we set. And we’re probably going to set it about a 70% level of predicted accuracy, which is still pretty high, 70% level they get. The rest of those patients will move to confirmatory NGS next generation sequencing testing, whether they have the mutation. But we’ll be able to do this at all much earlier in the journey, find them a lot faster, move them forward into the program, and actually then, you know, have it be. And for them, I mean, this in itself is life. Now, if this works, probably want to add this as a surveillance layer rate and standard of care. Now take what I just said. Now go to an area like let’s go to countries that are gigantic health systems like India or something like next generation sequencing is really expensive.
Jeff Elton: Same price there as it is here in the United States. In fact, samples come to the United States oftentimes for India based patient populations. You could use this same example of different technologies on different cancer data sources with a high predictive accuracy recall at a much, much lower cost rate. So that would be, you know, dollars, maybe $10 as opposed to several, you know, a couple thousand or $4,000. So when you start thinking about these tools, you know, we we and we think about this all the time, we start thinking about how do you flex workflows in the health system at scale that allow you to constantly be looking for ways to improve it now? And partially the reason why we spend so much time on the R&D side is that’s where you can get the validation of things and begin to apply the things. But then both we and Caris can move it over into the decision flows on the health care provider side. That’s. I really liked your explanation. And in our world, if we can constantly go back and forth on that, that’s how you make a real difference.
Harry Glorikian: Well, it’s sort of like an end around, right? I mean, you and I both know they the sequencing should come much earlier in the process, but if you can actually see features that identify, right, that this is that subpopulation that could that should go for that next sequencing as confirmatory that that’s huge….And this is the funny part for for those people that are listening, you know, Jeff is sort of like I mean, he’s sort of standing still because now, even after these things I’ve talked about, you still have Memorial Sloan Kettering, PathAI, and Bristol-Myers Squibb. So I don’t know if you want to say anything about those, but, I mean, I don’t know, Jeff, over the last two years, I don’t know. I mean, it seems like not much is happening over there. Right? As we were.
Jeff Elton: We’ll get you some other things in January, too.
Harry Glorikian: I mean, it’s funny you were joking earlier. You’re like, I’ve been so busy, I haven’t had time to get in trouble.
Jeff Elton: I was as we were coming on, I was saying, uh, you know, normally trouble had been able to find me, but I’m just too busy for it now, so, uh. Uh, so, um. Yeah, uh, maybe I’ll pick a high level theme. So health, health data, and R&D, you know, questions are they’re always complicated. Right. So you named a PathAI collaboration, Caris collaboration, Memorial Sloan Kettering partnership collaboration. Um there’s a reason why we work with, you know, the best minds, technologies and other companies in the world, because these are really complex problems. And, um, we we care about speed. We care about, you know, there’s again, you know, probably most of the people that work in my company, we’re we’re we’re we’re 1200 people now, um, a lot of people come from, uh, cancer epidemiology, clinical engineering, cancer data science background. Everybody came here for a reason. If you had my chief revenue officer on. He’s a PhD in cancer biology who has a head of commercial who’s a PhD in cancer biology. And when he would say he’s kind of quoted and said, you, you have me over a barrel. I could never work for another company because my whole life is dedicated to cancer outcomes. And it’s like, you know, like this is like the place, you know, so, so when we work with somebody like Memorial Sloan Kettering, they develop a lot of their own AI models that they apply for their own interpretations that are not developed commercially.
Jeff Elton: And you remember the world of laboratory developed tests where you could dom there’s the same sort of principles. And radiological AI where you can develop locally AI models and deploy it for clinical diagnosis and interpretation. And so they’re probably one of the more broad users and deep users of some of these tools and solutions. If anybody in the world, they are the world’s largest cancer research and cancer care site, and cancer, uh, they they have about a thousand clinical trials ongoing at any one point in time. So we’re going, you know, we’re working with them to take our Terarecon tools, develop the research version of those solutions that do AI model development at scale for deploying both in their workflow for research purposes, ultimately go up into care and diagnosis. And so all of this, these are all accelerators, take the best minds and work with the best people to actually take really hard problems and work them through with speed, alacrity, and get things done, get it into place and kind of pull it through. So that’s, that’s our that’s kind of our, our kind of, you know, mantra about things.
Jeff Elton: And we and you know, everything’s hard. Right? So these are all like demanding people and the best people. It’s tough. And you have complex and challenging, and sometimes I would even say the conversation could have an unpleasant feel because somebody’s challenging what you did. But that’s also part of the process of getting to something that truly is better and worthy. And, you know, that we can kind of depend upon and get everything out there. And, you know, you may know if you go around, you can look at our website and you look at our ASCO, we publish everything. And I want to get back to your AI. One of the reasons why we publish is we have to be a high trust entity. Right. So we innovate so rapidly that by the time something gets published we’re already on the next generation. So we came to the conclusion that protecting intellectual assets actually had lower value than working the assets, getting trust in the assets and just out innovating and just kind of, you know, by the time something published, you’re you’re already on the next generation. So it doesn’t matter that they knew what you did before because, you know, that informed what you’re already working on now. And those are also principles about how we operate.
Harry Glorikian: So just to sort of put a wrapper around this, I’m thinking like, if you zoom out and you look at the big picture here. Yeah, I wonder if it’s fair to say that the story of progress in medicine, including drug discovery of new drugs and the design of clinical trials, is now all coming down to data how much cancer data you can get, who controls it, the quality of it, and how you can use it to train, you know, these new AI models. And I think all of this was probably true even before generative AI, but now it feels like it’s even more true and more relevant. And so I’m, you know, want to ask you, like, do you agree with this idea that data is now king?
Jeff Elton: Um. I think that our ability…I’m going to link a couple themes just from even your earlier questions. And I want to take your question super seriously because, you know, it starts to say, how do we really operate? How do we set ourselves up, and what should you provide access to? So the level of cancer data that we now have access to that are derivable out of every instrument, machine, etc., is really unprecedented. And what we’ve now seen when we leverage that, particularly when we kind of take it from multiple cancer data sources, that that multimodal concept at very large scale, the insights we can generate with causality, with kind of conclusive understanding of something, is unlike anything we were able to get even five years ago. So when we think about how we’re going to practice research and practice medicine. And I really like the way you asked that question earlier, because I actually think how we practice research, how we practice medicine actually may methodologically become a lot more alike, which is back to your data question. So I actually think it’s super important that cancer data are made available. It’s super important that we’re careful about how we use data. And we actually kind of, you know, kind of constrain, but it’s super important. And that was one thing we learned during the pandemic, too. Going back to your pandemic, more data became available from more sources because we needed to move quickly.
Jeff Elton: And we saw when we do that, we’re able to actually crack the back of some really tough problems really quickly, too. So I think one, all the data element you said absolutely correct, but also now it’s the methodological tools that we have and the processing capacity and all the new cloud infrastructure, which is also evolve at the same time, that infrastructure that powers GPT 4.0 and that powers some of the stuff that Meta is doing and the stuff that Nvidia is doing. It is extraordinary. So when you pull that set together, then I 100% think that’s how you know, we’re going to be doing. And the thing is, the data is there. But what we know is still a very small fraction of what’s knowable. Right. So this is where, you know, if you if, you know, hopefully if you think I’m excited about this, I’m wildly excited about this. I mean, you know, to the point where it’s like, you know, I’m like every day I want to kind of see, like, you’re almost going to feel like can’t possibly go fast enough just because the progress you can make is extraordinary. So we even challenge ourselves, like, can you change how you operate to go faster? So anyway, that’s that’s that would be my response back to it.
Harry Glorikian: So I mean, it really seems, and I could be wrong, you know the numbers better than I do, but that nobody in oncology world is wrapping up more cancer data than you guys right now. So I mean, do you feel that that the companies in a position to sort of be the dominant player in clinical trial design for new cancer drugs.
Jeff Elton: Um, so, uh, I think we’re in a great position to be a leader in doing that. Um, I want to I want to come back, you know, whenever I think of the world of data, if you saw my internal publications and even stuff I do with my own people, I always express a reverence for the patients who had the experience that generated the data that we are a responsible custodian of. And I’m trying to pick my language very carefully here. And yes, we by far are in the best position of doing that. But we also realize the obligation, I mean, you know, the privilege and the obligation that comes with that. And part of it is our responsibility is to use that position in all the right ways. Why would we work with an ASCO to make sure that whatever you need for any research that you actually believe is going to be beneficial to patient population, it is made available to you. We’ve never charged a not for profit university or academic researcher for a cancer data set, ever. And we never will. And we provide them for PhD theses. And, you know, whatever the case may be. So yes, we’re I believe we actually have the highest quality data. I think we have the most rigorous standards. And I think we have better access to anybody else there. And we want to be the partner for everybody doing the best research and the best activity. And we’re actually going to make sure we work it to actually provide the best care for people, making the best and the most effective tools for that, while recognizing all the other, you know, kind of other responsibilities that come with that privilege.
Harry Glorikian: Well, I always tell people I mean, move as fast as you can because I’m not getting any younger. So, you know, I want to benefit.
Jeff Elton: On that one. Man. We’ve been doing this for a while. Yeha. I’m hoping this keeps me young, so. Well, just.
Harry Glorikian: Yeah, yeah, yeah. So, well, Jeff, it was great getting an update from you. I know, I mean, so much has changed. It seems like. I mean, not just ConcertAI alone, but all the technology and everything that’s also helped drive it. Right? You need you need a synergy of of activities to, to help drive a company forward like this. But, um, you know, depending on what you guys do there, you could be a third time on the show could be the charm.
Jeff Elton: Yeah. So, Harry, I really appreciate you’re always I always enjoy your conversations. You always ask great questions. And it kind of forces me to be thoughtful. But I think you’ve been in the industry and you know the right you know the right domains and the right areas. You know what matters yourself. So I really appreciate the opportunity to be back here again.
Harry Glorikian: I look forward to maybe running into you at JP Morgan.
Jeff Elton: I’m sure we will. All right.
Harry Glorikian: Thank you.
Jeff Elton: All right. Thanks, Harry. Have a good one.
Harry Glorikian: You too.