What if there were a single company that could connect hospital electronic health record systems to a massive genomic testing and analytics platform? It would be a little like Amazon Web Services (AWS) for healthcare—an enabling platform for anyone who wants to deploy precision medicine at scale. That's exactly what Joel Dudley says he's now helping to build at Tempus.

When Harry last spoke with Dudley in January 2019, he was a tenured professor of genetics and genomics at the Icahn School of Medicine at Mount Sinai Medical Center and director of the Institute for Next Generation Healthcare. But later that same year, Dudley was lured away to Tempus, founded in 2015 by Eric Lefkofsky, the billionaire co-founder of Groupon. 

Tempus is building an advanced genomic testing platform to document the specific gene variants present in patients with cancer (and soon other diseases) in order to match them up with the right drugs or clinical trials and help physicians make faster, better treatment decisions. In this week's show, Harry gets Dudley to say more about Tempus's business—and explain why it was an opportunity he couldn’t turn down.

You can find more details about this episode, as well as the entire run of MoneyBall Medicine's 50+ episodes, at https://glorikian.com/moneyball-medicine-podcast/

Please rate and review The Harry Glorikian Show on Apple Podcasts! Here's how to do that from an iPhone, iPad, or iPod touch:

1. Open the Podcasts app on your iPhone, iPad, or Mac.

2. Navigate to The Harry Glorikian Show podcast. You can find it by searching for it or selecting it from your library. Just note that you'll have to go to the series page which shows all the episodes, not just the page for a single episode.

3. Scroll down to find the subhead titled "Ratings & Reviews."

4. Under one of the highlighted reviews, select "Write a Review."

5. Next, select a star rating at the top — you have the option of choosing between one and five stars.

6. Using the text box at the top, write a title for your review. Then, in the lower text box, write your review. Your review can be up to 300 words long.

7. Once you've finished, select "Send" or "Save" in the top-right corner.

8. If you've never left a podcast review before, enter a nickname. Your nickname will be displayed next to any reviews you leave from here on out.

9. After selecting a nickname, tap OK. Your review may not be immediately visible.

That's it! Thanks so much.

TRANSCRIPT

Harry Glorikian: The last time I had Joel Dudley on the show in January 2019, he didn’t sound like a guy who was looking for a new job. At the time, he was a professor of genetics and genomics at the Icahn School of Medicine at Mount Sinai, and the director of the Institute for Next Generation Healthcare. He was publishing breakthrough papers on the use of advanced statistics to find unexpected biomarkers for diseases like Alzheimer’s.  And he had a long to-do list of ways he wanted to push his fellow physicians to become more data-driven.

But lo and behold, later in 2019 Dudley was lured away from Mount Sinai by Eric Lefkofsky, the billionaire co-founder of Groupon. Lefkosky had started a new company called Tempus, with the goal of creating an advanced genomic testing platform to help oncologists and other physicians make faster, better treatment decisions for their patients. 

Lefkofsky showed Dudley what the company was doing to document the specific gene variants present in each cancer patient, in order to match them up with the right drugs or clinical trials. And it didn’t take him long to talk Dudley into joining as chief scientific officer.  In our interview, I got Joel to say more about why joining Tempus was an opportunity he couldn’t resist.

One cool piece of news that came out right after we talked is that Tempus isn’t just a provider of testing and genomic analysis—it’s now a hardware company too. This year the company plans to release a portable, voice-driven gadget called Tempus One that will allow doctors to interact with Tempus’s genomic reports through natural language inquiries. It’s like Siri or Alexa, but specialized for oncology. I’ll have to get Joel to come back to tell us more about that. But for now, here’s our conversation from early January.

Harry Glorikian: Joel, welcome back to the show.

Joel Dudley: Thanks for having me back.

Harry Glorikian: So, you know, as we were just talking before I hit the record button. It feels like when we last did this, it was almost a lifetime ago. Especially the last few years, it  feels like, every day feels like a month, almost, trying to keep track of everything. But, you know, you were doing something very different the last time we talked to you. You were at Mount Sinai and and now you're, you know, at Tempus. And so let's start there. Like, why the switch and. What are you doing?

Joel Dudley: Yeah, I think, like many people, I didn't expect to be at Tempus. I've been here about a little over a year and a half now at Tempus, and I was approached by Eric Lefkofsky, the founder of Tempus, when I was at Mount Sinai. And things were going great at Mount Sinai. I was fully tenured. I had tons of grant funding, cool projects, even startups spinning out of the lab. So I definitely wasn't looking for a job at all. And and I hadn't really heard of Tempus at the time. And I just knew they were kind of out there. And I somewhat heard of him and he approached me about a job. And I'm like, yeah, I'm not looking, you know, and I know Guardent. I know people at all the sort of big precision, Freenome, and precision medicine companies. I mean, I thought, well, if I was going to go, why would go to Tempus. You know, and like, I just, I know everybody else in these other companies. So he's like, just come to Chicago, you know, talk to me and see what's going on.

Joel Dudley: And then I looked at the website and I'm like, how the heck is this company worth three billion dollars, you know. $8 billion valuation now. And I'm like, I was being, to be honest, a bit arrogant because I'm thinking I know everybody in this field and I don't know what these guys are doing. Which is a little arrogant, to say that. But it's like sort of like, how could a precision medicine company get to $3 billion without me knowing about it. So at that point, it was almost curiosity at that point that brought me into their headquarters, obviously back when we could fly and travel. And I went I went in there. I'm like, well, I've got some collaborators at Northwestern anyway I've got to meet with. And yeah, I'll just go I'll go see what this tech dude wants. And I was even telling my wife before I left, I'm like, all these tech guys, they, always have the worst health care ideas, like, they have the worst health care ideas. 

Joel Dudley: So so I'm like I'm like, you know, but that being said, I went and visited Eric at headquarters, Tempus headquarters. I was completely blown away, completely blown away. It was a company like nothing I had ever seen before. And I can get into some specifics on why Tempus was different. But at a high level, it was really the first time. So my background, I'm very much a systems guy. Right. I like to understand everything from multiple systems perspective. Right. And in the molecular world, that means I'm a systems biology guy. I want proteomics. I want genomics. I want the whole thing. So when I look at other companies that were doing targeted DNA panels, I'm like, well, what fun is that? You know? And I know there's a good reason why people do that because of reimbursement and and all that kind of stuff. But it's like, what am I going to learn from DNA? You know, nothing. So that was my bias. And Tempus was the first precision medicine company operating at scale I saw that was totally committed to a multi-scale multimodal data philosophy, which I had never seen before, and was totally committed to this concept that I think you and I get excited about, which is a diagnostics company that was first and foremost a data company, first and foremost. Now, there's a lot of diagnostic companies that paid lip service to being data companies. But when it came down to it, there were all about volumes and margins of their tests. Right. Tempus was the first one that was authentically and seriously and in a big way committed to being a data company first.

Joel Dudley: So I was totally blown away and and at first, you know, said there's no way I'm leaving my great job here in Mount Sinai. And I kept thinking about it and I kept thinking about it and I thought, holy cow, these guys are successful. This is going to be massive. I mean, this is going to be bigger than anything I could do at any single academic institution. This is going to be world changing. So anyway, that was a lengthy explanation of why I joined Tempus. It just wouldn't get out of my brain.

Harry Glorikian: Well, it's interesting because I remember when you told me, I was like, what? Huh? Like, I was adding up what you were adding up, like all the different things you're doing. And I'm like, he went there? I'm like, I almost was thinking, can I buy stock? If he's going there, I should buy stock. So you know, Eric, before he did, you know, Tempus, obviously, did Groupon and, you know, he's financially successful, I could probably say. But what was his motivation?

Joel Dudley: Yeah, he the origin story of Tempus is that Eric's wife had gotten breast cancer and someone of great means, of course, was able to get, have her seen by all the best, literally all the top the top 10 cancer, breast cancer doctors in the country. And what he noticed, being, if you get to know him, he's a very rational, logical guy know, very data driven guy. He noticed very quickly that, you know, first of all, none of the doctors agreed. That data wasn't informing her care, you know, and got a real personal look at sort of the dysfunction, I guess, or let's say missed opportunities to use data in health care that we see we, you and I see. And he decided to do something about it. There's a lot of really admirable things about his personal involvement in Tempus that drew me there. One is he's all in. I mean, he's all in, all in. A thousand percent of his attention is focused on the company. He's got a venture capital firm. He's got Groupon still is in existence and is in, and he is in in a huge way. He's you know, I think every time I've been to that office, I think he's the first one there in the morning. You know, it's just like, in some ways he's sort of like the general that rides the first horse in the battle on this thing. And not only did he not only was in a big way financially, he put a huge amount of his own money into into the endeavor, but his personal investment is, he's fanatical about Tempus.

Harry Glorikian: Well, I'm convinced that when you want to change the world, if you're not fanatical, then it's not going to happen. You have to believe it more than anybody else believes it to make it come true.

Harry Glorikian: Yeah. One of my favorite stories. I'll just share a quick note and I'll switch was I remember one time we were having a discussion. I can't remember what it was about. A flow cell, after I joined. A flow cell failing or something like that on the sequencer, and Eric I think had asked for which flow cells failed and I had walked by his office attempts and the bitmap images of the flow cells were up on his computer and he was staring at them intently. I have no idea if he even knew what he was looking at. I mean, he does now for sure. But the point was, the point was it was just shocking to me because I'm like, here's the CEO, billionaire CEO of this company, and he's looking at the pixel by pixel at these flow cell images, trying to figure out why they failed. And I thought that was unbelievable. You know, no, no detail is too small.

Harry Glorikian: No, you know, I think, you know, you have to be passionate, get involved and want them, you know, I mean, at some point you're at scale and you have to sort of start trusting the people around you. But in the beginning, you know, I think you have to fully be committed. And everybody has to be going with you. Yeah. So and I totally agree on the whole data driven part. I mean, I have given so many talks, especially with a good friend of mine, Jennifer Carter, who was the former CEO of N of 1, where, you know, there's a bunch of doctors where the genomic data is saying one thing and they decide to do another, which boggles my mind why you would do that, because most of the time it doesn't work. But so you guys are at the forefront of genomic data. And I'm sort of imparting words of saying, you're trying to get faster, real time patient care decisions and help physicians make better decisions. Is that, am I summarizing the business?

Joel Dudley: Yeah, yeah, that's it. In at a high level, it's obviously to deploy precision medicine at scale. So one of the things we say we're doing a Tempus is building all the boring, boring plumbing that nobody wants to build to actually deliver precision medicine at scale, which includes....So we ingest clinical records for the patients, because we contextualize the reports of the clinical data that we get from the individual patient. So but we work with everything from community, rural community hospitals to sophisticated academic medical centers. So we have this, part of our machine is, we have this interface that can take everything from a direct pull from a Cerner cloud instance all the way to literally people shipping paper to Tempus. But but, you know, basically we've built we built that data abstraction API, if you will, that can take eithr paper or cloud. And it was expensive. It required a lot of people and it cleans up the data. But somebody had to do that, like someone had to build that, the boring plumbing to do that. And and we did it.

Harry Glorikian: Well, Flatiron I think, you know, what I've heard is Flatiron has a bunch of people in the back end, like putting things in context right, yesterday versus tomorrow versus, you know, trying to get context, which NLP not very good at. And I got to imagine that Foundation might be doing some of the same sort of stuff. No, not as much?

Joel Dudley: Not as much on the clinical data. They're very much focused on the molecular data. The difference, though, between Flatiron and Tempus, though, is that Flatiron bought the EHR which the data was being collected. And so they own that. We take everything, like I said from manila folders to Cerner, to Epic to... Like that was the challenge, that's what makes TEmpus totally different in that we didn't own that that EHR. So it was a bigger challenge. But we also have humans that check all the data because as you mentioned, NLP is imperfect. But the real business, though, if I could make a point, though, is is developing smart diagnostics. Because, the principle being, you know, we all want to bring AI, let's say, to health care. One way to do that is to bring AI into the EHR, which doesn't seem like it's going to happen anytime soon. Like we have a hard time. You know, we barely can get logistic regression to run inside Epic. I don't know. I don't think we're going to, I shouldn't pick on Epic alone. But, you know, it doesn't seem like very sophisticated AI is coming to the EHR anytime soon. Plus, there's sort of a small number of players you have to deal with, you know, to have control over that environment. So that's challenging. You could try to bring the doctors to AI, which doesn't work very well. A lot of companies have failed because they say, oh, we have this beautiful AI machine, this beautiful interface that the doctors would just leave their, you know, standard workflows and just come over to our obviously better system. That feels like 99 percent of the time, right, because doctors don't want to change, physicians don't want to change their workflows. So the idea behind Tempus was more, physicians interact with lab tests all day long. So one step at bringing AI or a Trojan Horse, if you will, is to make the lab test themselves smarter. So a real simple example is, our cancer testing is, e because we pull the clinical data on that patient and the sequencing data, here's a real simple example of something that Tempus can do with a smart test that other people can't, which is if they have a DNA mutation that suggests the patient should go on a certain drug, but we know from their actual clinical records that they tried that drug and failed it, we will dynamically change the report to not put them, not suggest that drug or gray it out or whatever, depending on the version of the report. That's like a brain dead simple example, but most companies can't do that because they're not able to rapidly pull in and structure the patient's clinical data and contextualize the molecular data or the test result with that specific patient's information. So that's the Tempus approach there.

Harry Glorikian: Well, not not to not to digress, but I've always said in my talks, I believe that if anything breaks or will break health care, it's the EMR systems being completely, you know, I mean, they're just they're just not where they need to be considering how fast where we want to go to the next level of health care. Right. If we were a tech company, it would have been rewritten, you know, 15 times by now to get us to where we need to go.

Joel Dudley: Totally, totally.

Harry Glorikian: But you're looking at DNA, you're looking at RNA, you're looking and you're looking at a whole host of 'omics to help drive a positive outcome. I mean, are there concrete examples that you might give in how this is being used and why, you know, why Tempus is compared to everybody else where it is, I would say?

Joel Dudley: Yeah, absolutely. So you know what? One of the things that we think about when we get a sample in the door is how much sort of multi-scale data can we generate on the sample without going completely, without being totally insane. Right. So it's like I mean I mean, still being sustainable, let's say. So I'll give you. So what happens today when let's say, by the way, we're expanding outside of cancer, but focusing on cancer for the meantime, when a tumor section comes in to our current lab. So not only do we get sort of the the deep targeted DNA sequencing, we also get normal blood as part of that so we can do tumor normal. A lot of companies don't even do tumor normal. But then, and this is one of the things that really caught my attention, was, we generate full transcriptome on every patient that comes in the door. I mean, that's nuts. I mean, that was nuts that they just decided to as a default on every patient. That's like that's like $800 in extra cost that's not going to be reimbursed. And and even clinicians can barely wrap their heads around RNA today. I mean, it's a super hard time with RNA. I mean, do they like DNA because like the variant's there, or it's not, and the drug gets prescribed or not. But RNA is this analog probabilistic sort of dynamic measure. It gives you all kinds of different types of interpretation that's difficult. But the fact that they committed to that from day one was nuts.

Joel Dudley: So then we also have our own pathology lab. So we actually digitize the section and stain and digitize all the tumor sections. We have high quality imaging. And then we pull in the structured clinical data, of course. And then we have an organoid lab actually inside Tempus. So we try to build a patient specific organoid from every every patient we can and bank that for future screen. So we have a huge number of organoids where we have not only the organoid stored and the ability to really expand that but then the patient's actual, you know, in vivo clinical data, molecular data. And you could start to do things like, hey, where you know, if we if we see this pathway in the organoid, it means we're going to see this pathway in the real patient and all that kind of stuff.

Joel Dudley: So another interesting thing about Tempus is, we have this new business unit called Algos. And this is something that sounds really obvious when you pointed out and you wonder why nobody else did it. But we go to market with the broadest possible assay. So in a traditional, like, biomarker discovery, you would say, I want to try to find a biomarker of people who respond well to radiotherapy or something like that, prostate radiotherapy or something like that. So I'm going to start with the, people would start with their full transcriptome and then maybe, let's say you find a 10 gene signature that predicts who's going to respond well to radiation therapy. Then the the typical diagnostic company would say, OK, now let's shrink, let's take this 10 signature, let's implement it at Nanostring or PCR or some kind of care platform and and then go to market with that. And Tempus says, well, screw it. Let's go to market with the full transcriptome as our default assay, because then that allows us to digitally layer signatures on top of it. And by default, everybody. So we measure transcriptome now. And maybe five years from now, we find a new signature for drug response. We don't have to remeasure everybody. We just run it digitally, you know, on top of the signature.

Harry Glorikian: You know, that was one of the I remember when we were talking about this years ago, I was like, that's what you would want to do. That's why you'd want the data. Right. So you want all of this data so that as time goes on, you don't have to go back and get it again. You've got it. And you just look at it. It's almost like I think about it like topology. I mean, at some point you take the first scan and you start layering things on top to get a better idea of what what is there over time, because, hell, the technology, you know, your insight becomes better over time. Some new piece of information comes in, and you go, oh, let me go back and look at this again. So you guys do that. And then the recommendation is a targeted therapy. I mean, I haven't seen any of the reports, so I'm sort of guessing along here.

Joel Dudley: Yeah, we've got we've got a great report that summarizes the patient's clinical history and all the stuff you sort of expect. And then it offers various recommendations also about, of course, clinical trials. So the other thing we have is a huge clinical trial network, which I haven't mentioned yet. A national clinical trial network where we can spin up trials and match patients to trials. That's owned and operated by Tempus. But we can, so it takes the DNA information and RNA information and synthesizes recommendations. And it's going to be up to the doctor. Of course, you know, some doctors like to look at the DNA. Some people like to see where does the DNA and the RNA corroborate each other? You know, is there a PI3 kinase mutation plus activation or deactivations of a PI3 kinase pathway or something like that, and so we present all that information and a pretty, pretty digestible way.

Harry Glorikian: So, two questions. A, does the patient ever get something to look at? And B, have you done any stats on success, right, of recommendations and so forth?

Joel Dudley: Yeah, we've publishd some papers. We had a paper in Nature Biotech and a couple of, a couple of others that sort of show the value of this additional information and continue to publish, you know, papers. But we've been primarily on the cancer side, primarily physician facing. And, you know, physicians can, of course, give their reports to the patient's physician facing in other disease areas like neuropsych, which we've gotten into. We do have a patient facing digital app that is being tested right now to go more directly to patients, but not yet, and COVID as well. We have a patient facing up. So but that actually will be a bigger part of all the disease areas.

Harry Glorikian: You have agreements with tons of institutions coming in. I mean, you and I were at one point sort of throwing this idea of having enough data where you're at that escape velocity of, it sort of stops making sense to go someplace else because the Encyclopedia Britannica is in one place. So where are you guys on that journey?

Joel Dudley: Yeah, I think we're, you know, it depends. You could argue it, but I think we're basically approaching escape velocity at this point, where if you look at the trajectory of our data and I don't have the exact numbers handy, but it's a, it's a steep it's a steep line in terms of the number of samples we sequence. I think it's close to 200,000 samples last year or something like that. But but but our RNA, for example, our RNA database alone, I mean, the Cancer Genome Atlas looks like a little baby toy dataset compared to the Tempus's internal dataset. And that's, of course, a massive, I don't know if it's a multibillion dollar, but it's a massive Internet effort among academics. It's a great effort by the way, I'm not knocking the Cancer Genome Atlas, but but by comparison Tempus is able to eclipse that, you know, like you wouldn't believe. And then also have very much richer clinical data associated with those samples and have continuous updates of that data where something like the Cancer Genome Atlas is like this frozen thing that gets updated by an academic consortia every year. So even when we look at the cancer Genome Atlas, which again, I think was a worthwhile investment, and remains a worthwhile investment. But if you just compare those, the growth trajectories and the density and quality of that data side by side, Tempus is just a rocket ship compared to that data sets like that, which used to be like, you know, even Big Pharma would rely on the Cancer Genome Atlas is their sort of discovery data set. But now you'd be kind of insane not to use Tempus, it's just so much bigger.

Harry Glorikian: So so that brings me to that next question. Right. So we've got we've got these patient samples. We've got clinical data. You make a recommendation, you can actually recommend a clinical trial. But now the next step comes to me and says, well, but if I have all all those pieces of information, shouldn't I be also looking at drug discovery?

Joel Dudley: Yeah. So quick on the trial site. It's worth it. I'd like to point out 'cause we're really proud of this. So we have this thing called the Time Trial Network. It's a national network of I think it's 2,000 oncologists around the country on a common rate sheet, a common IRB. And the whole idea was when we match a patient, instead of a drug company going to, say, an AMC like Dana Farber or something, which, of course is a great institution, and saying, hey, we want to run our X, Y, Z drug trial with you, and all the patients will have to either fly here or drive here every couple of months, if you don't have all the patients here locally, we created this national network. And the idea was rapid site activation of trials. So if a pharma is looking for a certain type of pancreatic cancer patient subset and we match that patient in Tulsa, Oklahoma, or nearby or something like that, just picking a random city, that instead of that person driving into the AMC, an academic medical center that has the trial, or CRO, we spent a trial as close as possible to where that patient lives at one of our partners, whether it's a community hospital or something like that. At the end of the year, don't quote me on this, I think we had, we went from like a patient match to first dose in patient and something like less than 10 days or something like that, because we rapidly activate a single patient trial site.

Harry Glorikian: Wow, that's cool.

Joel Dudley: It's pretty cool. So it's sort of like a whole ecosystem. Right. So it's not only are we sequencing the patient and finding who are eligible, we can we also have the trial site integrated into our platform.

Harry Glorikian: So it it's interesting, you always wonder, like how much how aware our patients that some of these things are. Out there when they need it, right, as opposed to the way that you and I both know the way the system runs, which is, oh, come here so that we can make the dollars as opposed to what what's really going to be the best for the patient?

Joel Dudley: Yeah, yeah, absolutely. And you had asked me a second question that I totally forgot now because I distracted.

Harry Glorikian: The drug discovery side of it, making that connection at some point of...

Joel Dudley: Yes, it's super valuable data for drug discovery. And that is part of the value proposition of Tempus, of course, to our pharma partners who want to develop therapeutics. So part of Tempus's business is to partner with pharmaceutical companies and assist them in their discovery or biomarker efforts through Tempus's data and platforms. And we have some backend platform technologies for investment targeting our data. We have a platform called Lens for interrogating our data that is produced. Pretty interesting. And then, you know, we have a business called Alpha, which is about spinning out joint ventures around therapeutic discovery from from Tempus's data.

Harry Glorikian: Ok, so that's how you if you identify something, you're willing to sort of spin it out at that point and see it come to life.

Joel Dudley: Yeah. Yeah. So it's partnering with pharma or partnering with, you know, a joint venture that we're involved in around the data, but per se we don't do the drug discovery internally on the data.

Harry Glorikian: You and I love the data and love the AI and machine learning. What gets you super excited? Where do you see the biggest applications of the A.I. and machine learning? Where do you see the biggest opportunities?

Joel Dudley: And in no particular order, so a lot of interesting things can be done with machine learning when you have not necessarily orthogonal but multiskale data on the same samples. Right. So I'll give you a concrete example is, we have we have a large histo genomics, we call it program that our AI data science team is working on, where, of course, if we have rich RNA sequencing and rich DNA sequencing plus digital pathology on slides and samples, we can start doing things like calling PDL1 status directly from an H&E stain via deep learning instead of actually sequencing a patient. Because sequencing is great. But but imagine if you could call it the critical markers for a trial via an H&D stain and deep learning, you know, in rural Louisiana, or something like that, where people don't want to pay for sequencing or you just want to be much more capital efficient. So once we once we start collecting all these different dimensions of data, we can start predicting, you know, across all these different dimensions. Right. So what in the rich sequencing data can we predict from images, for example, which is really interesting, because then that cost, you know, nothing practically. But the key up front, you have to collect those those cohesive, coherent data sets of multiple dimensions to train. Once you've trained, it's super valuable.

Harry Glorikian: It's interesting because I was having a conversation earlier today about spatial resolution of single cell, but but actually looking at the genomics inside the cell, the expression patterns and looking at that based on geography, let's call it that, for so everybody understands it, but very cool how you could see individual cells lighting up versus, you know, the other cells around them, which would give you an indication of what's being activated, how it's influencing the cells around it, et cetera.

Joel Dudley: Yeah, absolutely. And that's an area we're exploring within Tempus, of course, is related to the histo-genomics I mentioned is if we start with a single cell and spatial transcriptomics on tumor cells plus rich imaging, at some point we're going to build up a data set that will give us deep molecular insights from the images alone, once we've built up the single cell and spatial transcriptomics that accompany those those images. So that's one, it's a really useful practical application of AI. Another one that's interesting for us is just getting additional insights out of existing data, which is something I've always enjoyed. But a concrete examples is, we have a big partnership with Geisinger where we've developed a deep learning model that runs on ECG traces. ECG traces are collected for elective surgeries, for physicals. And we're not the only ones necessarily exploring this, but a lot of people are using deep learning models to see if the, because an ECG trace, you could consider an image, basically. Right. And so people are using it episodically to see, like, is there something, that subtle pattern that's not being detected in the episode of care, but we're actually trying to predict things that will happen in the future. And we published some papers on this. But so we're taking a single ECG trace and we're saying, are there hidden signals basically in this ECG trace that will predict if someone is going to get future a-fib, future stroke future, you know, coronary syndrome? And we have a very large data set with Geisinger that we've done in partnership. And we've it's just amazing, like the one year, three year future events you can predict from a single snapshot of an ECG. There you go. Myocardia.

Harry Glorikian: Yeah, I like I have my little monitor here, and I, I, I tend to do it every day just just to get some longitudinal data.

Joel Dudley: Yeah. Yeah. Alivecor is a great is a great device. Yeah. So a couple of really interesting applications of that. One is, you know, from a population health standpoint, just going through all of the ECGs that have been collected and you can triage people into high risk low risk groups and manage them. But it's also interesting for clinical trials, because if you can predict things in the future from an ECG trace, say, for, like an anticoagulation trial, you can enrich that trial population for events and things like that from a fairly cheap standard device. So I'm interested in, you know, the ability of ML and AI to get additional, squeeze, additional information and utility out of these sort of everyday things that are measured routinely.

Harry Glorikian: Yeah, and I think that, I mean, you know, whenever I've seen it, we've always gone from a complicated measurement to figuring out easier modalities to sort of identify that information from. We just didn't have the, maybe the power per se to get it in the first place. So, okay, you guys are in oncology now, you're moving out to cardiology and I think infectious disease and do I dare say neurology, depression and things like that. So why? Like, why wouldn't you just go deep and, you know, crush the space in that one area? Why?

Joel Dudley: Yeah, it's interesting. I feel like we are doing fairly well in oncology. But this goes back to why I joined Tempus, which is, I always joke that this is like four different companies. And, you know, it's like it's like Flatiron plus Foundation plus, you know, we don't like to compare ourselves these companies, but like this is early on when I was, because we're actually not like those companies, which I'll explain in a second, but I was like, on the outside, it sounds sort of crazy to say, well, we're like six companies in one. But the difference was, it was built that way from the ground up in an integrated platform, a vertically integrated platform. And that's what makes it powerful. It requires a lot of capital to do that up front. But the vision was pretty interesting. So they built this sort of vertically integrated, very powerful machine to tackle cancer in this like multi-modal, comprehensive way. But they were smart in that they built it in a fairly abstract way so that it could be repurposed for for other diseases. And from day one, that was always the intention. And to me, that was amazing because I'm thinking, well, geez, a company that just tackles cancer alone with this approach is a massive company, you know,, putting on my venture adviser hat. You know, it's like, well, jeez, this is huge because this is like this company plus that company, plus that company all wrapped into one nice, seamless package. That's huge. And then I thought, well, if they replicate this success they're having clearly going to have in cancer in just one other major disease area that is an unprecedented precision medicine company in history. You know, no company would have done what Tempus has done in cancer and a whole other disease area in terms of ushering in this like very large scale multimodal approach, with clinical tests in the market and things like that. So I was like this, I got to join this. This is nuts.

Harry Glorikian: Well, it's interesting that you say that, right? I keep trying to explain to people and I guess one of the examples that I've been using lately is something like Ant Financial, right. Where how they started in one area and were able to broaden, based on some very simple capabilities. And now it's 10,000 people managing 1.2 Billion customers. Yeah, you don't do that because of a personal touch. You have to have automation to tackle that. And and I know that you guys have like your robotic systems for sequencing. And I have to believe that that thing doesn't, I always tell people it doesn't care what it ingests. Right. Analytics on the back end may need to be adjusted accordingly. But, you know, that's the power of this data approach as opposed to the way we've done it historically.

Joel Dudley: Absolutely. And the way I would describe it, I'm not sure everybody loves this analogy, but I think it's a very accurate analogy, which is, what I saw, and we're doing this, so we built this very sophisticated, vertically integrated infrastructure that connects sequencers to clinical and back, plus data abstraction and clinical data structuring. And so we built that machine and sort of dogfooded it ourselves on cancer and and other things that we continue to sort of dogfood it and use it our use ourselves. But eventually the goal of Tempus is to open this platform up to other people, so the way I what I saw early on was that while Tempus has the chance to become the AWS of precision medicine, basically. We're building all this boring plumbing or connecting hospitals. We're building this, like I mentioned, this API of data abstraction that can connect everything from cloud based EHRs to paper, you know, and everything in between. So at some point we want to open, and we are actually beginning some partnerships where we're opening up Tempus's platform, because if we've invested a billion dollars in that plumbing, then the beauty is, you know, you should is a startup. You don't have to do that now, just like AWS. You know, it's like now three guys in a in a garage to get out their credit card and start Stripe or Shopify or whatever the next big company is. And that was always been the aspiration of Tempus, not only to build this for ourselves, but to build it as an enabling platform for other people who would want to deploy precision medicine at scale, which is, we're actually executing on that vision in a serious way. It was more of an aspiration, I think, when I joined. But now we're full on executing.

Harry Glorikian: It's interesting. I mean, I remember you saying that to me, I want to say, last JPMorgan, when we were actually able to travel and sit down with each other. I mean, I talk to other people and I mention Tempus and some people go, who? And other people are who are very knowledgeable are like, well, I don't see what the big deal is. And so it almost seems like. Do you think people know what's there that they can take advantage of?

Joel Dudley: I don't think people fully appreciate it. And of course, there's a bunch of things I can't even talk about that are even more exciting that are being cooked up. But you'll be hearing about them soon. I think we'll make a few JP Morgan announcements, but it's sort of the M.O. Actually, one of the things that attracted me to Tempus was our CEO is very much a show don't tell kind of guy, to the point where even some people get frustrated because.. Nobody gets frustrated. But it's like, hey, we're doing all these amazing things and nobody knows about them yet. But but he's 100 percent right in that people will know when we're actually doing, once we're doing the stuff, right. You know, and and that was impressive to me because we're obviously in an area that's overhyped, you know, precision medicine, AI in medicine. And there's a gazillion companies out there doing proof by press release, you know, on all their vaporware. And Tempus is doing real, real stuff that's saving patients lives. And, you know, and they're being very disciplined about it and not overhyping it and just putting in the work. And then in the long run, people will know. I think it's going to be all one of those things, like who's Temples? To, like, Oh, my God, I had no idea, where did this come from.

Harry Glorikian: Yeah, and I think your biggest challenge is going to be the last mile, right? I mean, it's like Internet connectivity, right? Well, it's on the street, but how do you get it into the house? And the biggest complaint I always hear from everybody is getting this implemented at an institution is not trivial.

Harry Glorikian: I would argue that's what Tempus is mainly solving is that last mile problem. In fact, you know, I don't know how many institutions are connected inti Tempus, but it's well over 100 for sure. And that's a KPI that we're tracking. How much how many institutions we have last mile connectivity into. And that's been just growing up. That was a huge KPI for us the last last year. And it continues to be. But I would argue that's the problem solving, is that last mile, because we are in clinic, in EHRs, have bidirectional data feeds and decision support and a large number of institutions, it's just people don't realize it.

Harry Glorikian: Let me ask you to I don't even know if you're still doing this. You were part of the Institute for Next Generation Health Care. I don't know if you're still.

Joel Dudley: No, no, no. Not anymore. 

Harry Glorikian: OK, well, so I'm trying to get you to put your next generation hat on here for a second. And if you're looking at everything that's going on and where this is going, like where do you see the next big leaps coming? Where do you see the next changes coming in how we're going to make a difference for patients and hopefully bring down cost? And how is the technology that you guys are working on where you see it going sort of driving that next level of outcome for patients?

Joel Dudley: What I always like we always like to say at Tempus is we don't know, because it's actually it's a very Tempus-y thing, to be humble that way, because we don't know. Like. Well, we all we know is that, you know, we have to build this data set and we need to build these pipes and we need to, like because that will enable whatever the thing is that hits is the next big thing, I mean, clearly, like in cancer and other areas, we've got some clear value propositions and starting in cardio and neuropsych. But I'm convinced if Eric was on this podcast, the first thing he would say is, I don't know. We don't know. We do know that it's going to require huge amounts of data and we're going to, so we're going to collect that data and then hope we figure it out or someone we work with figures out what the next big thing is. But if I put on my my personal hat, I guess I've always been interested in prevention. It's not an area we work in at Tempus a lot, we work with a lot of late stage disease, obviously when you start in cancer, you're starting in some pretty heavy disease area, right. And life and death. But we are getting into cardiology and we're looking at endocrinology, diabetes. We have a big diabetes effort that will be announced soon. And so I think when the stuff we're doing in cancer or when the approaches we're building at Tempus can start to be applied to prevention, I think will be really interesting in terms of moving the needle. And then, you know, in post COVID, we'll see what happens with telemedicine. But right now, we primarily interface with the, and again, I'm speaking personally. I'm not divulging any any strategic roadmap or anything here. But I would imagine at some point if telemedicine continues to go the way it's going, there's no reason a purely virtual telemedicine company could plug into temper's in the same way an academic medical center does. Right. So which I think would would be enabling.

Harry Glorikian: Well, I would I would hope that that would be, I mean, if you think about the CVS-Aetna deal, I know that CVS, last year, you guys announced a deal with CVS, if I remember correctly.

Joel Dudley: Correct.

Joel Dudley: And so I think now that telemedicine has become much more. You know the way to do things, wy would you want somebody going to the ivory tower when you could plug them in through the system and interact with them there? And I mean, there's a huge cost savings. And and from a I mean, time standpoint, it's just more efficient.

Joel Dudley: Yeah, yeah, and we spoke with a institution which I don't think I can name at this point, but they had mentioned that during covid they had even spun up a tele-oncology practice, which was surprising to me because oncology is just one of the things where you think what's so complicated, you know, you can't spin up a tele-oncology service. But in fact, they had and and they did extremely well over COVD. And then when you start to think about oncology, well, it's like, OK, I mean, you've got to see your doctor. But then they're saying, well, go get your labs at Quest. Go get your infusion at the infusion clinic, you know. You know, it's not it's not like you have to stay in the doctor's office. And I started thinking about it. I'm like, OK, tele-oncology can work. So, you know, whether we'll see broad, you know, expansion of tele-oncology probably after people see the profits AMC made, or AMC but another health system. But so so yeah. So it could be even in oncology, we see totally virtual services, you know, plugging into something like Tempus.

Harry Glorikian: That would be interesting. I always think, like, I'm getting older. So the faster that we move into this new world, the happier that will be. I'll have a better experience, right?

Joel Dudley: Absolutely.

Harry Glorikian: So knowing the two of us, we could probably talk about this for hours. Right? Especially on the data side. You know, I think I think you're right. There's an under appreciation for where, once you have the data, what the different things you can do with it over time. It's more looked at from the science as opposed to the data side of things.

Joel Dudley: Yeah, yeah. And I think a lot of people who practice data science and machine learning know this, that it's just, huge amounts of data of high quality data just trump any, you know, sophisticated machine learning methods. What I mean is like choosing between like the latest greatest deep learning or whatever method, versus just having a simpler method with huge amounts of high quality, the high quality part being important, data -- I would take huge amounts of high quality data any day because that's way more enabling than whatever sexy machine learning method is. And it's usually the case that once you have vast amounts of high quality dfairly straightforward statistical modeling methods will yield just amazing insights that come as a virtue of the scale and the quality of that data. And I think that's the lesson I learned at Tempus is that data just trumps all from that perspective. Then I think it's important to point out, because there's a lot of tool-only companies in the field like, "oh, I got, trust me, this deep learning methd is better than that deep learning method. Or It's got this little extra thing. Or this topological method is better than deep learning." I's like, who cares when once you have the volume of data that we have?

Harry Glorikian: Yeah. The only place where I would not differ, but say, I think when you've got multiple high quality data sets, then you need a little bit of help making sense of it all, because the human brain was not designed to look at multiple pieces of data coming together and see patterns that it might not normally be able to sort of visualize.

Joel Dudley: No, that's absolutely true. And that's the and probably being oversimplifying that, because that's my career, has been multi scale data. It's like machine learning and stuff like that. So I feel like I should, yeah, that's a good point. But huge amounts of high quality data and this multimodal, you know, we always say multimodal, the multimodal aspect is really important because we want different high dimensional measures on the same sample or same individual, if you will. And obviously, longitudinal as a dimension is a very powerful dimension as well.

Harry Glorikian: Yep. Yep. No, well, this is something like, you know, I, I talk to people about and Joel, not to sort of build you up, but I mean, there's not many people that have the biological and the data background in one. We haven't I don't I don't believe we've graduated enough of them yet. We're moving in that direction, but not not enough of them yet. So it was great to have you on the show. I'm hoping that we'll actually get together sooner physically rather than later. But I have a feeling we're in this for another four or five more months. Before this thing starts dying down.

Joel Dudley: Yeah, probably, when we'll travel back, but it's wild. I was thinking, like I said, I maybe mentioned this last time. I've been at Tempus only like a year and a half and we've added five dollars billion of valuation in that time. But what's really cool about that is not that we're worth $8 billion in valuation because valuations are, you know, whatever, but is that there's a sense within Tempus that we are still a small, scrappy startup just getting started. So like that that's my favorite part about that number, is not that, because I think a lot of companies, if they had an $8 billion valuation they'd be like, "We made we made it. This is great." But Tempus is like, "just completely ignore that. We are just getting started." It doesn't matter to anything we do day to day.

Harry Glorikian: Well, I remember when when I was at Applied Biosystems, you know, the valuation was going off the chart because we were doing the genome. Couldn't install machines fast enough. And I remember talking to some of the senior people and saying, okay, well, what are we going to do next? And I remember the gentleman who was taller, way taller than me looking down at me and said, have you seen our stock price like we are? We're killing it. We're performing admirably. And I remember going home and telling my wife, like, I think it's time to sell some stock. Because that is not the right mindset for success.

Joel Dudley: Not the right mindset, no. Yeah, it's it's it's very refreshing, you know, that it's that attitude is just, you know, across the board at Tempus, everybody is like, we're just getting started. We're just getting started, heads down, keep cranking. And we really, you know, obviously comes from leadership, but we really block out any distraction that would come from from that type of valuation or whatever, you know. So it's really fantastic leadership on the part of Tempus.

Harry Glorikian: Well, one of these days, I hope to to meet Eric, he sounds like an interesting character. But you know, stay stay safe, stay healthy, and, you know, obviously, you and I will constantly continue the conversation in the background, but is great to have you back on the show. And you know what, honestly, huge change from Mount Sinai, I never thought you would leave that place, considering.

Joel Dudley: I never thought either. But I enjoy it. It's been, like I said, as I've been recruiting people, I said, you've got to, like I don't care how good your job is now. You've got to get out now. There's like there's this wave where, everybody's going to be riding in the next decade, when I talk to someone like me. You're so well positioned to do it. And you're going to, if you don't get out and just try, you're going to kick yourself in five to 10 years and say, I saw this coming. I saw this big thing coming and I didn't get out.

Harry Glorikian: Well, I've been saying, you know, since we since we were doing the genome. I remember telling all my friends, I'm like, "Biology, man biology and where the data is going is where it's going to be." And people were like, "Well, tell me specifically where to put my money." I'm like, look, I'm not, I can't tell you right now specifically. I'm just telling you that that whole area is going to explode. And I think it's just going to, I mean, now we're at a point where it's, the curve is ridiculous. Gene editing stocks. What's happening in the space. I mean. COVID has pulled stuff forward in a way that I could never have imagined.

Joel Dudley: Yeah, me either. Yeah. Yeah, it's a huge catalyst. I agree, though. It's amazing. Good good time to to be in the field for sure.

Harry Glorikian: Oh, best job in the world. I always tell people.

Joel Dudley: Yeah, yeah. Science fiction is a cool business.

Harry Glorikian: Oh yeah, yeah, yeah, yeah. You got to have a little bit of both. Otherwise it gets boring.

Joel Dudley: Yeah, exactly. Awesome man.

Harry Glorikian: All right. Good to talk and we'll stay in touch.

Joel Dudley: All right. Sounds good. Take care man. Good to see you.

Harry Glorikian: All right.

Harry Glorikian: That’s it for this week’s show.  We’ve made more than 50 episodes of MoneyBall Medicine, and you can find all of them at www.glorikian.com forward-slash podcast. You can follow me on Twitter at hglorikian. If you like the show, please do us a favor and leave a rating and review at Apple Podcasts.  Thanks, and we’ll be back soon with our next interview.