The Pomp Podcast - 304: Jo Bhakdi On The Future Of Genomic Sequencing

Episode Date: May 28, 2020

Jo Bhakdi is the founder and CEO of Quantgene, a company deploying the world’s leading Deep Genomics solution. Their mission is to save lives through early detection, better prevention and more effective cures for all disease, starting with cancer. In this conversation, we discuss machine learning, sequencing technology, and DNA extraction procedures defines the cutting edge of genomic diagnostics, early disease detection, and precision medicine. ========================= Pomp writes a daily letter to over 50,000 investors about business, technology, and finance. He breaks down complex topics into easy to understand language, while sharing opinions on various aspects of each industry. You can subscribe at www.pompletter.com =======================

What's up, everyone? This is Anthony Pompliano. Most of you know me as Pomp. You're listening to the Pomp Podcast, simply the best podcast out there. Let's kick this thing off. Joe Bakhti is the founder and CEO of QuantGen, a company deploying the world's leading deep genomic solution. Their mission is to save lives through early detection, better prevention, and more effective cures for all disease, starting with cancer. In this conversation, we discuss machine learning, sequencing technology, and DNA extraction procedures, defining the cutting edge of genomic diagnostics, early disease detection, and precision medicine. I learned a ton from this conversation and I hope you enjoy it. Don't forget that I write a daily

letter to over 50,000 investors about business technology and finance every morning. I break down complex topics into easy to understand language while sharing opinions on various aspects of each industry, you can subscribe at pompletter.com. Again, pompletter.com or go into the description and click the link there. All right, let's get into this episode with Joe. I hope you guys enjoy this one. Anthony Pompliano is a partner at Morgan Creek Digital. All opinions expressed by Pomp or his guests on this podcast are solely their opinions and do not reflect the opinions of Morgan Creek Digital or Morgan Creek Capital Management. You should not treat any opinion expressed by Pomp as a specific inducement to make a particular investment or follow a

particular strategy, but only as an expression of his opinion. This podcast is for informational purposes only. All right, guys. Bang, bang. Super excited to have this conversation with Joe here. So thanks so much for coming on and doing this. It's a pleasure, Anthony. Great to be here. Yeah. Let's just get started with your background. Obviously, you guys are working on some very cutting-edge medical and technology related products but what did you do beforehand to really get started down this path? Well my background to understand how I got to this point to you know build Quantene together with my amazing team is that I come from a very medical and bioscience family so I was born to you know my parents were both bioscientists very passionate

microbiology and other life science related things uh not in business just science um and yeah i grew up in a lab basically with my dad and learned about pcr machines when i was two without really understanding anything but i kind of saw it and then when it came down to uh choosing my you know college uh course um what what i would study i actually wanted to get out of that i thought like medicine i've seen that enough for 19 years of my life i like it it's cool but i don't know what else to learn here and so i basically decided let's become an economist i was very interested in business and also economics uh and so that led me down and kind of more the business path went into kind of strategy consulting then murder snake positions and a little finance

then quantitative finance and i found my passion for everything quantitative complex systems but especially fuzzy probabilistic systems, which are important in finance and in venture capital. And then it came full circle when a family member who's also a doctor and bioscientist asked me for advice on a big math and statistics problem in genomics for cancer-related things. And that was in 2014, and I really got sucked into that. I was at that time working on a quantitative venture capital framework with a slight focus also on life sciences on the investment side but that project kind of really triggered my interest because the deeper I dug into that question how can you profile DNA not fragments

but whole genomes and kind of DNAs from isolated potential tumor cells how could you determine if something is a tumor cell if you just look at the DNA which seemed like a trivial question back then But the more I dug into it, the more I realized this is a very complicated question. And I asked a bunch of friends at, you know, Berkeley and Stanford and all kinds of institutions where I knew they know a lot of things. And none of them could answer the question. And then we had some good ideas and they said, well, that's definitely not something anyone has done so far. And we kind of solved a very important math statistics question in genomics. and that triggered basically the foundation of the creation of quantene because i realized well

if you can solve that problem and no one has solved it we can detect cancer at early stages in the blood because then we had the tools to basically identify specific locations on the genome necessary to detect cancer comprehensively in the blood and so that's the whole thing so So maybe let's go deep into the science here because I think a lot of people listening, they recognize that they don't understand the science nearly as well as you do. So help us understand when you're talking about DNA, like just what is that to get started with? And then we can maybe talk about how certain cells are mutated or kind of die through disease and how disease really kind of runs through the body and what that does to the bloodstream and to those DNA cells. sure so as a first step I think it's always important to see what cancer detection and cancer is today and it's an enormous problem 600,000 people die

every single year in the unit the United States alone of cancer 1.6 million get diagnosed with cancer each of us has a 40% lifetime chance of getting cancer it's pretty severe and the 10 over 10 percent chance of dying of cancer so it's probably one of the largest risks each of us has. And early detection of cancer is a huge game changer. If you look at survival statistics, the difference between, for example, breast cancer detected in stage one versus stage four is very massive. 99% of women have a five-year plus survival at stage one, 9% survival at stage four. So it's a massive difference, you know, what your chances are because if it's detected early you just go into a hospital get surgery and you're

cancer free in many cases whereas if it's safe stage it's like a or it's terrible it's an ordeal that goes on years and years of chemotherapy and very bad outcomes so you want to detect it early now genomics is something that was not used or is still not really used in in cancer detection it is something where we do hereditary testing more and more which means we look at your healthy genome right so all your cells have a genome a dna and that is kind of the building blocks of life are proteins and the proteins are encoded in the dna so the dna is being read out turned into proteins and we have like roughly 20 000 proteins in the human body different ones and that determines who you are now that is called hereditary you know genetics so all your cells have the same dna we

can look at that DNA very easily nowadays and then tell you if you have certain predispositions right if you have certain variants that other people don't have you might be more protected or less protected against certain cancers so that's the first step of genetics into cancer that we can tell you if you have for example BRCA1 variants specifically once in as a woman your your chance of breast cancer is much much higher that you get that over lifetime because something in your cell and metabolism is less likely to protect you. And how do we find those variances, right?

So like we know that now if you have that variance, you have a higher probability of getting breast cancer for a woman, for example, but how are those discovered? You mean how we scientifically discover the initial insight that there is a problem or how do we find it in your body? The initial insight that that is an indicator. Right. Yeah. So there are multiple ways of doing that. There's a mechanistic way where you basically identify a variant and then you try to find out what that actually does. So you can see,

if that variant is on the DNA, how does the protein change? And you can say, well, if the protein changes and that protein, for example, is a tumor suppressor gene like TP53 and it's being changed, then it's a loss of function effect, for example. Then the protein stops working. And if that protein is designed to kill a cell, if it becomes a tumor cell, right, like a guard that is in the cell waiting for something to go wrong, and that guard becomes disabled, then we know, okay, that will probably lead to cancer. There are lots of other examples, like there are tons of things that you can do. That's one way of figuring it out. Another way is population health studies, where you have basically association studies. So you

look at you know millions or thousands of people and you see well uh in the subgroup of people who have that variant breast cancer is a hundred times more prevalent like often it's not a hundred it's 10 times or something um and that then it's more you know correlation versus causation of one knows exactly what's going on but you derive from that well probably bad news if you have that variant We don't know exactly what it does, but you're in that new bucket, and we see that people in that bucket have 100, 300, 1,000% more chance of developing a disease. So you should be cautious and be on the watch out. Got it. And really what this gets at is something I find fascinating personally, which is just we really don't understand the human body that well, right?

we understand a lot, but it is such a complex machine that we're still learning things every single day. And so we're kind of doing the best we can with the information we have, right? And that's exactly why we built Quantine and Serenity, our kind of patient-facing brand. I think there's a big misconception about science and medicine and multiple misconceptions. But one of the biggest one is that we know a lot like first of all that's always a dumb assumption we don't even know like you know we don't know a lot about anything physics or chemistry this is all like the all these scraping the surface if you're lucky but in biology it's especially extreme I think when you are

in that field like every single time you scope below the surface you say oh this you know causes cancer or something and then you dig deeper into the protein and the protein change the number of proteins we actually understand what they do in cancer is probably maximum a handful probably more one or three like k-res or you know p53 or something i mean even there it's very unclear all the details are extremely unclear and then you dig one layer deeper and you have only questions and no answers and you know every good scientist would tell you well we simply don't know like you have to run a bunch of experiments once you run the experiments you probably still don't know and you only have some statistical

probabilistic insight and i think that's very important in medicine it's very different from for example building rockets right anything that's kind of artificially created by humans we have a much better understanding of if you build a computer or a rocket you have a pretty comprehensive understanding because you build it everything that's naturally occurring you have just no clue like when you look at the body if you're lucky you know like 0.01 percent of what's actually going on and it's very clear you know the deeper you dig they're just vast areas of unknowns yeah and so would it be fair to say that in a lot of uh the work you guys are doing uh obviously it's one trying to solve some of this but but two just in the broader uh science and health industry

is that uh we're really focused on just trying to understand what's going on right right the the solutions only come after we understand and so that is really kind of the the base layer and a lot of this biotechnology and things that are being seen as breakthroughs really is just identifying something that's been going on in humans for you know thousands of years we're just now starting to understand given the technologies that we have absolutely and i think there is a huge paradigm shift going on in medicine right now very slowly going on so that's why we created a company to accelerate that but the paradigm shift is the medicine of the past and the current medicine the way of doing medicine is is very static right you go to med school you

learn a bunch of things and you think you know something which is a big mistake and I think the new paradigm in medicine that you know serenity and quantity stand for is that it's a much more dynamic system that is an equation of basically three things data you need a lot of data you need intelligence you need to make sense of the data and then action or systems right you need to then take some form of action and have some kind of infrastructure to do treatments or more diagnostics that's not how medicine works right now medicine right now static knowledge you're like if that then do that whereas you know I give you one example if you do a conventional

blood test, PSA, for example, for prostate specific antigen, you get one number. It says, oh, seven. And they say, oh, it's above the threshold. I might have a problem. Maybe not. When we run a test on you, especially the somatic mutational test for cancer detection, which we didn't get into yet, that's a different thing. We are not looking at your a healthy genome um we get six billion data points so we get six billion individual nucleotide reads out of that sample and then we have a very complicated uh cloud-based you know software and ai system that makes sense of that data and we are building bayesian layers on top of that to take in all additional information around you and that system isn't perfect but it gives you an idea

are you know what we are dealing with here that system changes every week like on on a deeper level because it learns from patients and studies it gives us literally six billion times more data points than a normal sample and it's operated on a different medical paradigm and the medical paradigm is more data is always better and more intelligence and maybe you as someone who's in tech and business you think like why is that like a thing like isn't that obvious I of course more data but in medicine you know the majority of physicians would say no that's not true more data is always often bad because more data means

I may have to make decisions that I don't want to make because if that test comes back positive I have to do some things according to standard of care therefore I would prefer not to do the test and for non-medical professions it sounds absurd especially for tech and finance like how is this makes any sense but it's about standard procedures it's about liabilities if i don't see it i can't do anything wrong and that sounds ridiculous maybe to you and many listeners but it's literally what most doctors would tell you like no it's a young patient don't do their tests because if it comes back positive i have to do all these things and that would be wrong to do because it's a young patient

and then you would argue but if you know it's wrong to do why do you do it it's like well because it's standard procedure and i don't want to get into trouble you know that's so let's just not measure anything it's absolutely wild how the health care system works it's it is wild and it's it's especially wild because you meet so many dedicated people who are high iq super trained and knowledgeable that do all these stupid things and know they are stupid and say i know it's stupid that's why i don't want to do it so it's this absurd system where we have so much talent and intelligence that is kind of yeah in chains they can't like they can't do what they should and want to do and that's why it's so important to break out and build companies that don't have

to adhere to these you know standard procedures but can really strongly innovate with the patient interest in time yeah and so let's talk a little bit uh about deep genomic sequencing right because i think that's a term that at least genomic sequencing people have heard before they everything from high level just like what that is all the way to they get to things around genomic editing and kind of all the scary stuff in most people's minds so maybe you can just start off with just like what is deep genomic sequencing and then we can start to talk about like once that's done all of the different applications of that information yes and that's exactly where the rubber hits the road so what i explained before hereditary testing that is the old school

sequencing that still hasn't reached most of medicine by the way so old school i mean it's five years ago it's still still only penetrated probably five percent of the of medicine so we need to do that that's very important but what that does is it looks at your healthy genome and to do that on a technical level um you know you're sequencing something with you know a depth of 10 to 30 x so what does it mean it means if you have a dna strand and you have a specific location you want to take you know 10 dnas and look at specifically that location and then you build an average and say it's a t t t t t g a t right so on the same location across 10 of these and then you say well an average is a t so you have a t right so it accounts for some sequencing

errors so you do it 10 times maybe you do it 30 times that's how you normally sequence your cells you basically sample a random 10 to 30 dna copies and look at the same location 30 times and tell you the average and say well anthony you have a t here and that makes you regular or it makes you different whatever it is now what deep genomics is and crunching is the global leader and precision sequencing so we have the most precise sequencing technology that can do the following if you have a blood sample and you have let's say 2 000 copies of dna in that blood sample of cell-free dna so 2 000 copies that stem from 2 000 different cells in your body that died and shed their dna into the blood our technology can make sure with a very high level of confidence

that we can investigate every single copy so instead of picking 30 randomly out and tell you what the average is we can tell you well we sequenced all 2,000 individual copies and for each of these copies we can say this is a T this is the T this is a T but here's a G and here's a T and so on and what that allows you to do that's a total game changer it allows you to say if there's any single copy in there from a cell that carried a somatic mutation. So a somatic

mutation is a mutation that is not normal for your body. That's not your healthy DNA. It is something where a body cell that you have is mutated, has changed. Very different from a hereditary variant. So it might be that if you have

blue eyes or green eyes or something, you have a G where other people have a T on that location. But all your cells have that. That's why it's you. Whereas what we are doing is we want to identify of all the DNA, cell-free DNA in a sample, is there a single copy that stems from a single tumor cell, right? And that gives you a level of precision that is completely paradigm shifting because before in cancer detection, you have protein tests. They need literally millions or billions of proteins in that sample to see any kind of delta, right? If you have a PSA of seven versus one, I can't tell you the exact number, but we are talking

about millions to billions of actual PSA proteins you need to have in there to see that difference. Would it be fair to say, as you're looking at the blood of an individual, would it be fair to say that you're grabbing the dna cells out of that blood and when you're looking if you're only using a sample so let's say 10 or 30 or 50 or whatever the number is uh you could actually miss one of the mutated cells because it doesn't end up being in your random sample so by actually um sequencing all of them you're not only one becoming more accurate but two it sounds like this precision technology you have actually lets you test every single one of them right so it's kind of precision leads to higher degrees of accuracy? Exactly. So it's very simple. The math

is simple. If you have a tumor in your body, early stage, right? That tumor at stage one has 100 to 200 million cells. That's stage one tumor. That's somewhere in your body. Out of these cells, a certain number dies every day. Actually a very significant number, right? So if you have 200 million, at least two to 3 million actually turn over per day. So they shed their DNA into your blood if you take 20 milliliters of your blood that's uh roughly you know 0.4 percent of your blood so it's not non-trivial right it's nearly half a percent of your entire blood if you take and so you can ask the question if you know you have let's say to be conservative a million of these tumor cells die every single day and shed their dna into the blood of course it gets

digested so you have to divide it by 24 hours and then see how much is actually circulating it every minute but you still if you do the math you still get to a point where it's very likely that you have 10 20 30 uh circulating two more dna copies in in one tube um so for us but you have thousands and thousands of other cells in the tube of cell free dna that you know messes this up so if you would randomly pick out you know 10 copies that's not enough you are very likely to miss anything in there if it's at that concentration so you have to make sure that you look at everything in that tube and that is something even five years ago that sounded like complete science fiction we five years ago already anticipated you know where sequencing will be now so we knew it's

definitely possible and so we started developing everything so we can do it now um but 10 years ago i think 95 of scientists would have said that's complete science fiction that's like a work drive or something why did they why did they believe that though right so they said hey that's science fiction but was it because they couldn't actually identify all of the cells uh in the bloodstream was it something around uh their ability to scale the testing of each cell like like why did they actually think that that was not possible because back 10 years ago you know you need basically three things to really make that happen well four things you need some improvements or this like you you need some quantum leap inside advances in chemistry

or in strategy how you sequence so it was just completely not possible 10 years ago but then people found some cool things out about how you can actually barcode dna so to to reduce errors and sequencing that's one important thing that you needed an invention that we couldn't have made like someone came up with it randomly and without that it's not possible but it's also a quantitative infrastructure problem you need sequencing technologies that are thousands of times more efficient than 10 years ago and we just happen to have them now thanks to illumina and some other companies you need massive computing power that 10 years ago most bioscientists would have said that's just not possible to ever get there and guess what you know

things you know we have massive improvements in computing power we are reading like 200 gigabytes in the cloud in a cloud system per sample as opposed to one bit right if you i mean if you have a psa test and test seven no you need a byte i think for that so you need one byte and we we have 200 gigabyte files so just that delta you know you need multiple improvements on multiple fronts that are very massive and so i mean even in 2015 when we did the math on that um the costs for one sample based on the sequencing technology back then would have been roughly a million dollars per sample per sample yes and today you know we are far below a thousand. Yeah. And so help me understand, now that you're able to test every single cell that

you find in the blood and do it with an increased amount of accuracy, what does that allow you to identify or kind of do that for me as a patient or an individual, I either one, couldn't get access to before from an information standpoint, or two, can you tell me, hey, you've got a higher probability or you actually have cancer before any other test can identify it like how do you put this into application now that you have the technology to do it so very different from hereditary testing it just tells you what your probability is of getting something over your lifetime somatic testing tells you if you currently have a tumor which is of course vastly more actionable um so what it does is it creates very complex deep genomics patterns so what a pattern

means is we look at, you know, multiple tens of thousands of locations that we predetermined before. So we know if you look at all of them, you have very good chance of intercepting any kind of tumor variant across 15 different cancers. So for each of these locations, we then get what's called a frequency, right? Because it's not yes or no. It's like how many mutated fragments do we find and how many non-mutated fragments. So we see at that specific location, we have 0.1% carry a cancer associated mutation in your blood. At the next location, 0.56% carry something. So that gives you a two-dimensional pattern, right? A deep genomics two-dimensional pattern. And these patterns code for diseases. Of course, the question is for what

diseases and how do they code for it? And that's the big statistics and research question. So what we did is we are running very large patient trials. We have a 10,000 patient trial going on right now, 15 different cancer types, and a control cohort, so of non-diagnosed patients. And you have to build machine learning around that. That is also hard to build because you have to do many things differently from self-driving cars or recommendation engines. It's just a totally different problem. You have to build very different things. And we run your pattern against the historic data set so right we have your unique pattern and we run that against all the other patterns like pancreatic cancer early stage late stage colorectal cancer early

stage late stage and so on and control patients patients that don't have cancer but are smokers patients that are not smokers not cancer but are obese and so we we basically run that against all these different cohorts and get a match back and that match tells us well anthony looks most like x and second most like y and then you have a bayesian layer above that that basically says well now we know the genomics match now we also take into consideration your age your gender other background information we have to put additional probabilistic layers probability layers on top of it and then we get a final call that says well considering all these things the system thinks it's most likely nothing hopefully or it's most likely red flag for

early stage colon cancer which doesn't mean you have colon cancer but it means well there is something now to watch out and take potentially some action and what's very important for us is you have to not just develop technology here you have to think about medical practice because you can't just give that to a patient and say okay good luck you you have to say okay what do we do now and that creates all kinds of liabilities regulatory issues all kinds of things so what we are building is a system that solves the entire equation for the patient so we say well based on all these things we think you should do a stool test even though normally you were not supposed to do that because you have enough risk here that we say if we do a stool test like poop in a bucket

you know with some genomics test on top of it we can rule out colon cancer with a high probability or rule in right whatever it is yeah and so what i'm taking away from this is there's a couple of different steps here right so the first is what i'll call uh just collection of the raw material so you have to actually get the blood or whatever it is then there's an actual scientific process that is identifying the cells and kind of having that precision technology or science that you talked about. And now what you're talking about is this third step, which is really just data analytics, right? It's very kind of high level. It's very large data sets, but it's something machine learning type problem where you ultimately can say, hey, we have confidence that

we collected the samples from the people that we think we did. We've got confidence that we were able to use our precision technology to get accurate identification, kind of the raw data that we have, we're feeding to the machine is accurate. And then if those machine learning algorithms and kind of computers are structured correctly, they should be smarter than the average doctor, right? And come back and say, here is exactly the path that this person is on based on the genomic sequencing. And then you guys almost are going to build like a recommendation engine after that that then says you should go get this test or you should go do you know x y or z because of the machine learning has matched you to these other people and that's where we know

how to make that recommendation do i do i have that kind of right yeah i think you got very close there are two more steps that make this whole thing work after that the next step is clinical algorithms and you know that is something you cannot do through a machine learning system you have to do this you need a lot of people involved you need to apply what's called a delphi method so you have to ask a lot of experts for their opinion and what happens under all these circumstances right for all these different cancers how do we what downstream clinical diagnostics should be applied and here you cross the line from science into medicine which is a very different thing so medicine is all about you know who agrees who are other people saying the

same thing because you're you have to get to a consensus among experts otherwise you're screwed right you need liability out of the way you need to implement that so it is a very different field than just science the scientists can just say well i think this is the best solution because here's my statistics in clinical reality you have to say but who are the people who agree with that because it's a much more holistic problem and then the last step so forth is clinical algorithms that this whole thing needs to work with which is very complex that's one of the main reasons why smart tech guys who go into biotech often fail because they lack you know the understanding of what goes beyond science and engineering what goes into this weird clinical space you know this

authority and doctors and experts and legal and compliance but you have to solve that otherwise you don't have a product and the final step is of course the business model like who actually pays for that and that again sounds maybe trivial to most tech and finance guys like why is this i mean if you save people's lives they just paper it guess what not in medicine so what well we change that fundamentally but uh it gets insanely complicated how do you convince a insurance bureaucrat that he likes that and if you show that you save people like sorry to be so frank but they don't care right it's not their job you have to show them that they save money and you have to show them that you save money immediately or quickly over time you know and then you know

things get insanely complicated and that kills also most of the innovation so if you don't solve four and five all your amazing tech is not going to work because you can't finance it And so this is like the, now I gave you the full picture and complexity why it's absolutely non-trivial to truly understand your business model, to innovate on that front, and to be highly synced with the medical and clinical reality of things. And then nail the tech. And then you're getting somewhere. Yeah. No, it's absolutely nuts. And I guess part of that clinical side, when I hear the word clinical, I think of all sorts of different tests that can be run with this genomic sequencing stuff.

Is this only being done on humans? Was it previously done on animals? And then that's where the technology kind of was perfected. And now it's being brought to the human side. Like, what is the relationship there in some of that testing to non-human kind of clinical trials or whatever you could do there? well the problem is or the advantage of genomic sequencing is you don't have to do a lot of animal trials

because there is no risk in just collecting the evidence right so that was very human driven from the outset on Craig Venter and the human genome project so you can sequence you know a lot and you don't you know without any kind of approval because if you

do clinical trials you just sequence no one gets hurt as long as there is no feedback so it's very different The actual sequencing and developing sequencing technologies, you're free to test this on humans because you're not putting anyone at risk, especially if there's no feedback of information. Once it gets into cancer screening, of course, the game completely changes. And that's the problem.

That's what's holding the liquid biopsy industry back. So what we discuss is called liquid biopsy because it's liquid and kind of a biopsy. um and that's where we innovated a lot to actually get out of that problem um and and get medical advancements and technologies to our members much faster than we normally would be able to do that if you just go through medicare and you know commercial insurance got it and then i guess as you um get better and better at this right because part of the machine learning algorithms that are actually analyzing all of this sequencing uh they should improve over time right it's kind of the beauty of machine learning. Help me understand, is that something where, let's say I'm testing for

a variation that we believe could lead to a higher propensity for cancer, right? Can I use the advances and the accuracy that we've determined on that test in those machine learning algorithms to then apply to a different type of test? Like, is there some shared knowledge or shared analysis that can be moved from test to test? Or is each one of these tests, when you go to start them, they're really starting from scratch, right? So you have to rewrite the algorithms and kind of start all over again

without any of that shared knowledge. I mean, it's one of the, I'm super excited about these things because it's probably the most challenging and rewarding data science problems I've ever seen. Well, there are some in finance that are also pretty immediately rewarding, but I mean, in a human sense.

and you can view it as like in a way if you take a step back it's like any data problem right if you want to determine something if you want to figure out if a certain real estate you know offering is like likely to increase in value over time you could build a crazy machine learning thing around it like how many people look at it where it's listed you know what the current price is what the stock market does like you could feed all that information and how does the stock market data relate to that real estate as opposed to other real estate listings around it well no one knows you can these are two different tests right you combine these test sets and you

know if you have sufficient data sets you can make them work together and that's if we test your hereditary uh if you if you understand your hereditary variance and we understand your somatic variance the somatic variants are also like a time-based uh kind of on a timeline time series because we do that every year whereas whereas your hereditary you don't have to do every year because it's not going to change um so you have very different types of data sets and so you have to learn how to make them compatible um but each data set adds a significant amount of insight and dimension to the problem um i think there are two major challenges in medical data sets the first one is they are often very heterogeneous so if you

have a thousand or ten thousand patients the probability that you have the same data sets for these 10,000 is like zero unless you do a super clean clinical trial that costs you an immense amount of money if this is commercial patients and I might have your hereditary genetics but not the next one next persons I might have you hereditary new somatic and good clinical records but for the next person I might only have hereditary and some records so when you build machine learning systems it's very important to account for that I bet you a sample the comprehensiveness of data is merely impossible to maintain and

just cutting it down to this to the data dimensions that are comprehensive is kind of stupid because you're losing so much information so you need much more fuzzy you know systems that allow you to recognize much more things and you also need to embrace more you know man machine hybrids where you have experts and analysts sitting in front of these things and use the machine learning more as a decision support system and have these mutual interactions it's very important medicine and then the ultimate problem of course from a business perspective is you need to close the business model loop in order to even get

that information right you need to figure out how do I actually make money and create a business model here and that's very very hard and I'm very happy we solved it the way we solved because for most companies that means you have to just do very rigid clinical trials over a decade that costs you hundreds of millions of dollars and are super inflexible and in the end you have bad data sets because you missed out on the most important element. We see this with a bunch of our peers that just don't have the right data sets. How do you guys make money what is the business model that you figured out that works so we made a very bold

step that some love and some hate um and we said i'm an economist by trade right so i love to look into a system you know and understand why is this not working and medical innovation innovation is simply not working if you see what's possible in theory and what's happening there's a huge delta between what could happen based on all the tech we have and what does happen in medicine it's ridiculous my conviction after looking at this for like over a decade now is it all boils down not to the fda not to the government not regulations that's smaller obstacles it boils down to the payer problem if you have socialized medicine which we effectively have or bureaucratized medicine which means if i protect your life as with a new technology the value i'm generating

is your value because you value your life and you know your health and your family's health but you're not the payer right or at least not normally we change that so if i have to create value for people but get paid by other people i'm i will stop creating value for you because you're not paying me i start to creating value for the insurance company or medicare which is the same thing so what do they want they want to cut costs they want to collect your premium and pay as little as possible right and they want to do so on a short time horizon they are not interested if i save the money in 10 years because i prevented you from having stage 4 cancer and saved you know hundreds of thousands on your cancer treatment they ask me if i give you that

test now within three years can you recoup the test and of course i can't it's not possible because you know if you test 100 people you detect two cancers for these two you save a lot of money maybe but maybe if you wouldn't detect it maybe they would have had symptoms only four years down the road so they're off the you know table for insurance companies so you know once we address only the payers interest the patient is out of the picture no one is going to innovate anything that saves you a life or protect you because you're not paying and the payer is not interested in that it sounds very brutal but it's also very obvious for everyone who knows business and so what we said is if you want to build a company that stands for true medical progress

you can only do that if you have a self-payer model you have to get paid by the people you protect because if you let other people pay for it you're not going to protect them anymore um and so what we develop is like can we develop something that is affordable for most people but still provides margins for us that makes us a real business and so we are like you know roughly 200 a month membership fee for once a year tests it's 2400 a year for that deep somatic test it's an 800 one-off fee for the hereditary testing but that's a once in a life uh thing you do then we have an intelligence layer once we do your genetics testing that you continuously get updates on any drugs that are incompatible with your genome that pose a risk to you any other

insights we have on an ongoing level that's 20 dollars per month ongoing so it's all affordable but over time of course we want to make this cheaper and over time this allows us to actually built the patient base and the clinical evidence to get insurance more and more on board because we're developing that evidence over time but not by raising a billion dollars and you know do a 200,000 patient study over 10 years so it's a little controversial to just go for safe payers but I will be see two things the market trend is on our side because a lot of patients understand that you have to do that if you want to get actually advanced medicine and second

You know, we also see that it's the only way of doing it. Like, as I said, as an economist, I have no illusions about incentives, you know, incentives in business. If you just want to save costs, you're not going to save people. It's very simple. It's funny how the same capitalistic tendencies that the finance industry has exist in medicine as well, right? I mean, yeah, to think, oops, yeah, to think that giant publicly traded companies, you know, pharma, insurance, and so on, don't think about the bottom line is a little delusional. And, um, once you understand the logics of, uh, you know, they're all, it's not that they are bad people, but you know how it works.

You have to report your quarterly profits. You have to show that you grow your business. Same with pharma. If your business is to sell chemotherapy drugs and you make $161 billion a year as an industry, are you really incentivized to cure cancer with like vitamin C? I'm not saying that this works, but I'm just saying, let's imagine vitamin C would work that would be the worst disaster in the history of pharma that actually works they would lose 160 billion dollars a year and so the incentives are a little flawed right so no one

is gonna i'm not saying anyone's actively trying to not cure it but if i pitch any big you know pharma company with oh i have this great idea it costs only 10 bucks to cure cancer if it works i need 10 million dollars to do this trial versus stage one trial i mean you can be guaranteed that no one is going to finance it because they say that that's an absurd proposition to us yeah it's pretty crazy um the other thing around uh genome sequencing that i think a lot of people um you know have heard of but they probably don't understand uh is when you get into all of the gene editing right so whether it's at the individual cell level or kind of this dna editing maybe talk a little bit about, it's kind of hard for me to see that world happening if we don't actually

understand first, right? We kind of have to understand the genes and the DNA and kind of what's going on before you could even think about editing. But there's obviously technologies out there that people kind of have heard of at a high level. And how do you think about that as maybe kind of a long-term path that the medical community is likely to pursue, whether it's just for curing diseases and kind of direct health applications versus like, hey, I want to change the color of my baby's eyes type stuff? Yeah. Well, I think the extreme depth of potential knowledge and the lack thereof in biology sounds like a daunting problem, right? So, oh, we don't understand anything. How can we even edit? But I think this immense amount of

depth and complexity and the lack of knowledge opens the door also to genius, right? You can be very genius by saying like maybe we don't have to know all these things maybe we can hack it like for example if you know a mutation does some bad stuff but you have no idea what you just know in the outcome is bad but we cannot figure out what it actually does maybe you don't need to figure that out maybe you just remove that mutation through gene editing and you know maybe you cut it out and replace it with a healthy piece of dna so the whole crisper revolution i think it's it's, you know, often things get overhyped, but CRISPR is a little bit like the internet.

You know, everyone said the internet, oh God, it's going to change our lives forever. Everything is going to change. And everyone's like, oh my God, so obnoxious. These people, they are overhyping this. But it was just true. I mean, it is changing everything. And CRISPR is very similar

because CRISPR allows you to, it's kind of crazy when you think about it. It allows you to, in a very targeted way, take any specific segment of the DNA and replace it. with any desired fragment you want to put in there through the nucleotide you have 3.3 billion nucleotides on your dna they can to the single nucleotide say okay at that specific location cut it out and inject that sequence and if you and they can do it in living organisms they can basically inject it into you and it would kind of you know disperse across your whole body in

theory and replace it in all your cells which is extremely freaky because it actually works right so you could actually change your eye color in theory while you're like living and or even change all kinds of things so that is extremely freaky of course there's an un you know enormous amount of things that can go wrong you're probably going to die if you want to change your eye color because some stuff is going to happen that no one you know was considering so i wouldn't test it out at home um but in theory it's only possible and so we know a lot of things but compared to what we should know it's basically zero but that doesn't mean you can't do absolutely amazing or shocking things and uh of course i'm very you know we

are very focused on disease and keeping people healthy and you know not not let them age too fast um there are many things you could do but of course you need very very deep clinical trials for that to prove safety especially if you start editing genomes and living humans that's it's more regulatory issue i mean scientifically it's definitely extremely possible and not even that i mean you can do it today and where do people get so like let's say i want to take one of those nucleotides and i want to replace it right so i've identified that that one is a bad one whatever for whatever reason where do I get the replacement is that something that I take from somewhere else in the body is that something that's like quote-unquote lab grown where does

that replacement one come from in that scenario so here you know and these are great questions because if you go into biotech and you will learn about these things it's absolutely stunning you know how hard certain things are where you think that can't be that hard and how easy certain things are where you think this must be completely impossible. And then someone's like, no, it costs five bucks on the website and you get it. And it's like stunning.

And this is one of these things. You can literally say, here's my sequence, T-T-A-A-C-G-A-C-G-A or something, right? That you, I don't know where you get this from, but whatever you do your research and say, I want this little part of the protein change, these amino acids.

And so I have, you know, these 24 nucleotide sequence that I want to inject here. You can literally go on the website. I'm not going to promote any specific companies, but you can Google that quickly. Copy, paste your T-T-A-C-T things in there, take out your credit card, and they send it to you two weeks later.

No way. In a little tube. No way. And then you have them. And they are clean. You have a few billion of these little sequences in there, and they're all lead sequence. And what are people doing with those today?

Are they like injecting them? no there's a huge amount of stuff that where this is used we do this every day we order them every day so um you know you use you need them in sequencing for example for pcr reactions so they are called you know these nucleotide sequences are hugely you know used all across the board they are for example used as primers so we can use them to basically have a little sequence that reflects another sequence on the dna and if you put that into a mix and do some stuff with it heat it back up and down and put polymerase in it this enzyme then these primers will anneal like they will align to the complementary the complement on the dna

and then start reading that out like whatever comes after so you can basically say i want to read out whatever is on the dna after that sequence that's actually sequencing so you say okay if you want to read out that specific part of the dna what you have to do is a primer that you synthesize that sits here and it then aligns here and then the polymerase enzyme comes here and starts complementing the rest of the dna and then that gets put on a sequencing machine and you read it out and you see oh yeah this is my primer i recognized it that's what i designed and then what comes after is the sequence that i was looking at so i can see if you actually have a t there or a t it's kind of similar to like in the movies when they uh they like clap really

loud so they can sync the audio and the video right it is kind of very similar yeah it's the same thing kind of yeah yeah um before uh before we go to wrap up uh you have been a a bitcoin proponent uh and have a pretty cool story uh about uh chamath who's uh who's obviously come on the podcast uh maybe tell a little bit just about like how uh you discovered bitcoins i think it was just pretty interesting that uh yeah you had some very cool people on the podcast i'm a big fan of chamath i think he's uh of course super smart but he's also outspoken and he is what a smart investor always should be and a smart entrepreneur he has his different perspectives he thinks in first principles about certain things of the world and then buys them same with kathy

from art right also amazing so i met chamath in howard business school there was an event i didn't go to how they were just at the conference and i think in 2014 and uh back then i had i had heard of bitcoin but not really i mean that is six years ago and he gave this very you know passionate pitch about bitcoin and how it's going to change everything and he you know he made this i don't know he just totally convinced me to get into bitcoin just with a little bit of money so I bought a bunch of Bitcoins at $800 because it's how I do my investing, right? I analyze stuff,

but I'm also really listening to entrepreneurs and smart investors where I, on a meta level, see, okay, this guy knows something. He's smart. The way he thinks resonates with me. It's the way I think, but he thinks about different things. So I'm listening to him on these different things.

And that paid off pretty well. So thanks so much. yeah that's an awesome story um before uh before we wrap up i always ask two questions and then you get to ask me one to finish it up but uh what is the most important book that you've ever read well i actually forgot the title of the book it was in german i'm from germany so but that book what was actually the title i know the cover was black and had some weird orange shape on it i read this when i was 10 11 or something and that book was basically taking you on a journey i forgot you have these you know there are these famous guys in the u.s who do the

similar thing it takes you on the journey on the evolutionary journey from the big bang to us today and i remember one specific thing in the book that was kind of a game changer in my mind and that was the theory of life right so the most exciting thing for me is not how monkeys turn into humans it's like trivial i mean a monkey is basically a human or like even a fish turns into whatever a rat kind of that's like fine i get it so but the question is how does this all start how does something dead turn into life and in this book that is he described that in a way that was so logical that was shocking how dead stuff turned into life that basically if you have a bunch of molecules randomly assembled proteins right in the in the soup

the beginning soup thing right in the sea and you just put energy and and light on it so they start changing they mutate randomly not as life but as just dead matter as proteins like oh yeah has a little light here and it changes like that at some point randomly in that soup a specific protein or substance or molecule will change in a way that it will align to other molecules and replicate itself and that's just a pure question of statistics you to that point like of course at some point that's just a random trait of the thing but at the second way it does that guess what happens to the soup these molecules would be much more you know prevalent

than all other molecules, because they're the only ones who replicate. And that's the first step to life. That's what viruses do, for example. So viruses are not alive, in case people didn't know. It's not a life form. And so, you know, that was for me stunning, because it was exactly that magic steps,

like, ooh, there's dead stuff, stones, and there's life, when you look at humans and stones. But when you narrow it down to this one thing, what is actually the main trait of life, to understand, well, the only thing that it does, it replicates. And, you know, and that is purely, that's pure math, statistics, chemistry. It will happen inevitably because whatever replicates will be more of the other stuff. And then it's just the race who replicates more effectively, who can eat other stuff, who can develop DNA. Then you're off to the races. But that was such a revelation because it's a purely mathematic explanation why life is inevitable. I remember the first time, and I forget the name of the book as well, but I read about telomeres, and they were basically talking about the replication. And you're just like, wow, this is a whole world that I don't understand. I don't know if I want to learn more because it's scary. But at the same time, it helps you understand life for sure.

uh you spent a lot of time thinking about uh human life obviously and kind of what humans are made of and dna sequencing and all that uh what about aliens are you a believer in aliens not believer what's the thoughts there um well the fermi paradox is one of my favorite topics like why you know if you do the math on it it's kind of obvious we should have millions of alien civilizations even higher developed civilizations and why didn't we meet them it's one of them you know are they just lingering out around earth and don't make contact because we are too primitive or what's the thing and i actually i have this i have my own theory there but i think it must be true that there are tons of civilizations um but it also is very likely

that they don't go beyond a certain civilizational development stage um which allows them to travel everywhere like quickly and the question is why so i think these are kind of first principles how i think about it there must be all these civilizations they must be highly developed and it's extremely likely that nearly none of them makes it beyond a certain point which is a little scary when you think about it and that point is we are not at that point it's a point where you could easily travel everywhere and see everything which in theory shouldn't be a problem right so my theory is it's a little dark normally i'm very optimistic but i'm just reasoning here it's not my opinion i just reason like what's most likely

and absurdly it has actually something to do with something very trivial you would think and that is terrorism i think the problem is if you develop as a civilization right there is a certain equation like how much power does this civilization have have like the power to change something and to destroy things it's always the same thing and of course there's a totally exponential trajectory you know if you go on this trajectory you have a billion times more power in a thousand years but that power also applies to each individual in that civilization right so if that power is so exponential it's only a question of time until you reach a technology point where every single person in that civilization becomes capable of destroying

the entire civilization like it's just a question of time and we think about atomic weapons and bioweapons, you take this times a billion, even if only 10% of the population would be able to destroy the entire civilization, you turn into that scenario where you have to rely on every single human being not to decide today to kill everyone. And I think that's an impossible equation. At some point, if you have a trillion people at some point, and every single one of at every minute could destroy all others what are the odds that this never happens like at some point becomes impossible and so i think exponential take and everyone talks about ai in the end i think that's a much more severe problem like at some point your power exponential curve

and the distribution the inevitable distribution of power leads to a scenario where you can't survive listen i don't think that you're very far off right which is kind of the scary part of a lot of this in that um the technology that we have you know access to as individuals today right people always talk about you have a supercomputer in your pocket compared to you know the 1950s or 60s right so that's only you know what if you kind of look at it's only 60 years of progress 70 years of progress seems like a lifetime right for for one human but on the grand scale of human progress it's a it's a blip so yeah no okay you just i mean here we are talking about why these civilizations don't exist and how they might get you know extinct if you take nuclear atomic

bombs or hydrogen bombs and compare them to whatever happened before what was the biggest weapon before like some cannon or machine gun or something in like 1890 so then you have an atomic bomb what's the delta it's pretty severe and now just go on another 200 years if that same leap happens again on a log scale the next weapon category then already becomes a little shaky the whole thing and then in another 200 years we have another log scale so you know you have something that's like 10 million times stronger than like a big nook so then you can already like wipe out the planet and then ai on top of that so i think it's inevitable that you generate more and more of these technologies that are extremely awesome and powerful but the margin for error

becomes just slimmer and slimmer like exponentially slimmer if anything goes wrong then all of us are gone and i guarantee you in a thousand years or two thousand years i mean there's no way around that little problem there might be a solution to the problem but it's a very complicated solution i see yeah for sure um to wrap up you could ask me one question what uh what one question do you have what do you enjoy most about your podcast and talking to different people oh this is the most selfish thing i do i get to talk to some of the smartest people in the world and ask them all the questions that i have right and yeah i record the conversations everyone else can listen but uh but for me it's just i get to learn um and by far that's the most rewarding and enjoyable part of it

is uh you know i learned things today i learned things yesterday i'll learn things tomorrow um and uh it's pretty cool that uh you know you would take an hour out of your time to sit down and kind of answer questions and and um and teach me things and so i'll keep doing how did you get into it what was your first step i'm kind of interested like how did you how did you engage on that journey and what was the main what were the main obstacles in the beginning that you faced yeah um there was a couple of guys who who came to me and said hey you should start a podcast and literally i was like well what's a podcast uh because i thought that i was listening to pre-recorded radio uh in the beginning uh when i would listen to things like a joe rogan uh whatever

um and uh we recorded three episodes and uh and we launched it and they were a huge part of helping me kind of figure out you know how you launch a podcast and all that kind of stuff and uh there was just enough people who paid attention to those first three episodes where i said well maybe i should do a fourth or a fifth and then kind of from there it's one of these things where uh you just kind of put one foot in front of the other right and next thing you know you turn around you're like oh i've done over 300 of these now like that's kind of a lot right um and so uh it's been fun and i'll keep doing it until uh one day i'm gonna wake up and say all right no more fun i'm done like i'll just walk away i don't know when that's coming but uh but i don't think any

time soon awesome yeah i thanks a lot for doing that i always like you know i admire people who built these podcasts because i learned a lot from you know for sure where can we send people so that they can learn more about you and in quantum um choose serenity.com uh that's our product site So choose serenity.com or directly to quantgen.com. And people can also reach me at jb.quantgen.com if they want to get in touch directly. Got it. So jb.quantgen.com and we'll put some links in the description for you as well. But listen, Joe, I really appreciate you doing this.

This is super fun. I learned a ton and I hope everybody else does as well. Awesome. Thanks so much, Anthony. I really appreciate it.