00:00hi everyone welcome to the a 6 in Z
00:02podcast I'm sonal and I'm here today
00:03with Gabriel OTT who is the co-founder
00:07and CEO of free gnome which is a company
00:10that aims to cure cancer through early
00:11detection using machine learning and
00:13other software techniques on DNA we also
00:16have our general partner Vijay Pandey
00:18who heads up our bio Fund and we also
00:21have malinka wali oh my god emoji or
00:24last name wallah Lee update perfect and
00:26that's so embarrassing because I'm
00:27Indian too and I really should know how
00:28to say your last name it took me a while
00:30so well great you guys welcome we're
00:35talking about machine learning
00:37techniques brought to by o and what that
00:39opens up for us and what some of the
00:40challenges are as well so just to kick
00:42things off yeah I mean one place to
00:44start is that you know ever since the
00:46human genome project people have been
00:47talking about all these great cures and
00:49things I will come from it and what is
00:51different now you know why are we
00:53expecting to see something now I think
00:55one of the biggest realizations of the
00:57human genome project is that the human
01:00body and the genome is much more
01:01complicated than we thought everyone
01:03thought we would know the cause of all
01:05diseases and therefore we would be able
01:07to cure all diseases as soon as we
01:10realized that 20,000 genes just were not
01:12enough to explain why all these diseases
01:15occurred it gave life to all sorts of
01:18new fields of biology that have been
01:20built around understanding the genome
01:22what's different now as opposed to what
01:25was happening in 2000 is finally the
01:27technology the machine learning
01:29techniques as well as the hardware
01:30supporting that has matured to a point
01:32where we don't have to try to manually
01:34figure this complicated system out by
01:37ourselves we can do it in a
01:38computationally aided manner which is
01:40greatly accelerating research in this
01:42video tell me why that makes such a big
01:44difference because you know my image is
01:46of genomic scientists sitting in a lab
01:48with like pipettes and just sort of
01:50doing like their studies or whatever it
01:52is that they do that's not in a horror
01:53movie sure I think in the 80's 90's you
01:57could get an entire PhD studying a
01:59single gene the simplest way of
02:00understanding genetic diseases is a
02:02point mutation in a single gene that
02:04causes a certain disease and
02:06Huntington's is a great example of that
02:08where a single mutation can cause a
02:11tremendous amount of
02:12nor the generation met at a certain age
02:15but most diseases we found out weren't
02:17like that and in fact most diseases were
02:19so complicated that we didn't
02:21necessarily know the root cause of it
02:22genetically still like just for a very
02:24simplified analogy if you think of the
02:26letters of the alphabet and just like
02:27words are created by a combination of
02:29letters you now have to actually hone in
02:31on a more overlapping set of variables
02:35is that overly simplified that is
02:36actually a pretty good encapsulation of
02:39what's going on there's an entire field
02:40called systems biology that's been
02:42created around understanding how genes
02:44interact with each other and yes there
02:47are the a C's GS and T's that's the
02:49basic code of life that creates a lot of
02:52these genes and any changes in that code
02:55could affect cellular function and
02:57therefore physiology level function but
03:00we realized that it's often a concerted
03:03change that results in a serious disease
03:06as opposed to a single change that can
03:09cause propagation down to the physiology
03:11level you know there's an irony here
03:13which is that in the old days people
03:14used to do things physiologically you'd
03:17look at tissue and that it was a major
03:19advance of microbiology would go from
03:21tissue to an individual protein or an
03:23individual and we'd go after that and
03:25only now that we're having this nuanced
03:27view that actually the physiological
03:29approach actually makes sense and that
03:30the system's approach comes in the other
03:33key part of what's now is that Human
03:35Genome Project cost three billion
03:36dollars and you know that's a pretty
03:37hefty copay for someone to have to shell
03:39out it's only now that genomics is
03:41starting to get actually to the point
03:43where you can imagine putting it in
03:44production where you would cost
03:46thousands of dollars maybe even hundreds
03:47of dollars why is that though I mean
03:49there isn't a Moore's law
03:50like there is an intimate sequel inton
03:53genomics yeah yeah there is I mean
03:54essentially it follows a Moore's law
03:56like curve the intriguing thing is that
03:58it's not the same hardware that is doing
04:00it behind it but it might be the same
04:02sort of wolf power and it's Moore's Law
04:05why several matters it yeah what is the
04:08it though like what is the thing that's
04:10accelerating the way Moore's Law and
04:12driving on the cost for biology the
04:14dominant player in the sequencing
04:17machine space has almost entirely by
04:19itself been driving down the cost of
04:22sequencing over the last decade or so
04:26think about genomics as a sequencing
04:27layer which is the companies that make
04:29the sequencing machines at Illumina
04:30which is like sort of Intel and chips
04:32and then there's the application layer
04:34which are companies that are using this
04:36genetic data to make clinical
04:37Diagnostics and that's sort of like
04:39Microsoft whatever people software
04:41companies building on the hardware so
04:42it's sort of like the equivalent in the
04:44semiconductor industry is the hardware
04:46and chip makers and the sequencing layer
04:49is the is the people who are doing this
04:52chips the fundamental substrates and the
04:55application layers of people who are
04:56building the software applications on
04:57top of it that we can use ok that's
04:59super useful so you guys have definitely
05:01answered why the shift is happening like
05:04a more systems approach things are more
05:05complex that's kind of obvious to me the
05:07complexity pretty much invites computer
05:10science as a as a point of doing things
05:12that humans cannot calculate but what
05:15are some of the manual vs. automated
05:16things that you guys described as then
05:18and now like what's changed in the lab
05:20and people's practice so a lot of my PhD
05:24work was in a field called computational
05:25biology which is in some sense a halfway
05:29house between completely generalizable
05:32machine learning approach to answering
05:34these questions and just looking at
05:36single genes computational biology is
05:39really a field that's designed to use
05:42computational tools aided by the
05:44biological knowledge of the scientist
05:46that's wielding those tools so some of
05:49the things that you can do in
05:50computational biology for example is
05:51take the DNA data that's coming from an
05:53Illumina sequencer for example
05:55tremendous amounts of data you're going
05:58to have to go through essentially
05:59prepare that data before you can do
06:01anything with it that step is called DNA
06:03alignment and that used to take days now
06:07there are tools out there that can do
06:09the alignment process in five minutes so
06:11that's some of the things that that
06:12field has contributed to in the last 15
06:15years where I think computational
06:18biology could use help from the things
06:20like machine learning is making sense of
06:23what that amount of data is actually
06:26saying about how we understand disease
06:29and how we understand human health and
06:31that's something that manually it's very
06:33hard to do because let's say you can ask
06:36what is the RNA expression level
06:39those 20,000 genes that doesn't answer
06:42the question of which genes are relevant
06:44for determining whether somebody has a
06:47certain disease or not it only tells you
06:49the raw values across those 20,000 genes
06:51then the human has to go in there and
06:53figure out what are the relevant parts
06:55now with machine learning we can start
06:57using computational tools to answer that
07:00question of what are the relevant parts
07:02of this data yeah in particular the old
07:04paradigm of looking at snips makes sense
07:06because that's what a human being can do
07:08you know we think even like back to my
07:10snips you mean yeah a single single
07:12point mutations so almost like back to
07:14thinking about Mendel and it's peas you
07:16know it's something where a human mind
07:18can understand that but that's where it
07:21stops and actually a lot of genomics
07:23today is just rows and rows of people
07:25looking at papers looking for these
07:27mutations and then using that to assign
07:29and you know biology most likely much
07:32more complicated than that ability
07:34that's limited by human minds and so
07:36that where machine learning comes in is
07:37that it opens up the door to analysis
07:39that really human beings couldn't do
07:40much like computation has done more
07:43broadly exactly a she reminds me of the
07:45podcast you guys did with Herman Niroula
07:47we were talking about simulations and
07:48improbable because what's interesting to
07:51me is this is not just a distant or
07:52mediation of existing tasks that people
07:54do were making it faster it's about
07:56opening up things that nobody could ever
07:57have done oh sorry yeah because of the
07:59way a computer works building off ML
08:01techniques allowing us to take and use
08:04much larger data sets than was
08:06previously possible in Diagnostics
08:07generally people have been looking at
08:09very specific parts of the genome in the
08:11past and trying to make interpretation
08:13space of that purely what gives able to
08:15do is take an agnostic approach and look
08:17across the entire genome and use lots
08:19and lots of different points of interest
08:21which obviously creates a lot more data
08:23to work with but because they use these
08:25very novel ml techniques they're
08:26actually able to accommodate those
08:28massive large datasets and make more
08:31accurate and more useful interpretations
08:32what's happened is there are certain
08:34genes p53 tiaras a trois EGFR that are
08:39sort of the usual suspects that's very
08:43something way worse like
08:47but anyway and and so the research has
08:51largely fixated around understanding how
08:53these mutations affect the cell affect
08:56the body where less research has gone
08:59into is what other aspects of the genome
09:02are actually involved in that process
09:04the genes that I mentioned we're really
09:06talking about less than one percent of
09:08the entire genome and so there is this
09:1199.99% of the genome that hasn't really
09:14been understood in terms of how they
09:17affect people's bodies and how these
09:19diseases come out from other parts of
09:22these right so how do they manifest
09:24themselves so this ties back so
09:25interestingly to the point you guys are
09:26bringing up earlier about the systemic
09:28approach why the singular focus never
09:30worked as much which worked before it
09:32doesn't work as much now because of the
09:34computer just at least has had its reach
09:36its limits I mean to be very clear
09:38there's been really really great
09:40therapeutics that have come out from
09:42understanding the cancer in this very
09:46focused way mm-hmm what sometimes
09:49happens when you are very focused though
09:51as you missed a bigger picture and what
09:54we're realizing is with respect to
09:56things like detecting cancer early that
09:59focused approach is not working there
10:02are aspects of it that's just much more
10:04complicated than we thought the other
10:05thing is that there's a differentiation
10:06between Diagnostics versus therapeutics
10:09so a therapeutic being focused that
10:11makes a ton of sense previous
10:12diagnostics were either something you
10:14could do by imaging or something that
10:16you do by biopsy and biopsy you can do
10:18genetics there but you're not going to
10:19do biopsies prophylactically I don't
10:21think you want anyone like taking tissue
10:23from all of your major organs like once
10:24a year that's a disaster really the
10:27other key advance here in the cancer
10:29space and genomic space is the fact that
10:30blood has so much genomic information
10:32even from things like tumors and that
10:34opens the door for these new
10:36technologies to come in I'm gonna have a
10:38really obvious question though which is
10:39why is cancered so hard yeah I get that
10:42it's a complex disease yeah I think it's
10:43a couple answers one it's not a disease
10:45it's really many diseases number two
10:47actually defining is the one I was
10:48studying biology many years ago you know
10:51my need reaction immediately after
10:53reading this stuff is like I'm amazed it
10:54works at all so the fact that apoptosis
10:57breaks down or there's some somatic
10:59mutations that mutations happen
11:01that's actually not that shocking I
11:02think the fact that actually our body
11:03does such a good job of maintaining
11:05things to me often sounds even more
11:07surprising but you know we can do more
11:09to help it in addition to machine
11:11learning be able to come up with the
11:12Quillen of new types of biomarkers that
11:14people couldn't come up with there's a
11:16learning aspect here of machine learning
11:18which is intriguing that as you get more
11:19data you get better and that's something
11:22that's really not like any other test
11:24where you know you a lipid blood test
11:27doesn't get better as you have more
11:28patients there and and that's I think a
11:30really intriguing aspect especially as
11:32we wanted to detect cancer early and
11:34earlier where there's fainter and
11:35fainter signs and especially for a wide
11:37range of cancers the question is why is
11:39it important because for some people it
11:42might seem obvious but it really isn't
11:45obvious the best drugs that we have
11:48today to treat cancer called
11:50immunotherapies only give you about 30
11:53to 40% chance of five-year survival in
11:55treatment these are the best drugs that
11:57we have chemotherapy and radiation give
11:59you less than 20% chance of survival on
12:02the off chance we get lucky and we
12:04detect this disease early for whatever
12:07reason that chance of survival goes up
12:10to 80 to 97% my god that's a huge
12:13difference we're trying to systematize
12:15that early detection so that it's not a
12:17chance event anymore that we can do this
12:20for everyone it's not about just
12:22Diagnostics curing the disease it's
12:24about the early detection and early
12:27therapeutics that enables the maximum
12:29chance of survival with that in mind I
12:33think a lot of us like to think that
12:34it's going to be the next greatest drug
12:36that's gonna cure cancer it's gonna be
12:38that you know silver bullet there isn't
12:40really going to be a silver bullet in my
12:42mind to treating a hundred different
12:44diseases simultaneously I'm glad you
12:47brought up because I take it for granted
12:48that there's this question of why it's
12:50important so why is the detection part
12:53early so hard mostly because we're going
12:56against the paradigm of medicine the way
12:58it's been practiced for the last 2,000
13:00years the way medicine has been
13:02practiced is we practice what's called a
13:04symptomatic medicine symptomatic
13:06detection essentially people have
13:08symptoms they come into the hospital
13:10they get diagnosed based on their
13:12symptoms and then they get treated
13:14the unfortunate thing with diseases like
13:17cancer and Alzheimer's disease and
13:19various other ages so ciated diseases is
13:21that by the time you show symptoms it's
13:23often too late so we have this paradigm
13:26of just studying symptoms and trying to
13:28match diagnosis to symptoms but in order
13:31to beat these these more complicated
13:33diseases these more virulent diseases
13:36you have to find a way you have to build
13:38a technology that can detect these
13:41diseases before the human being shows
13:43any symptoms and that often correlates
13:46which is very very difficult signals to
13:49find within the body yes there are
13:52signatures early on that indicate
13:55whether there is a tumor or not but what
13:58hasn't been solved yet
13:59is exactly what are those signatures
14:01that allow us to detect those diseases
14:03and what are the best signatures for us
14:06to detect them most accurately when you
14:09say early detection we should separate
14:10risk from diagnosis and so what about
14:13these bracha tests you know best cancer
14:15tests there's article about Angelina
14:16Jolie doing this some time ago so that's
14:18different because risk tests just tell
14:21you your chance of getting cancer they
14:24don't actually tell you if you have
14:25cancer right now and so if you had
14:27higher risk that probably means you want
14:29to get tested more frequently but we
14:31don't have good methods of testing and
14:33diagnosing you with cancer early right
14:35now yeah that's that's a great point
14:37we're not doing any kind of prediction
14:39we're doing detection it's not really
14:42empowering to the consumer to know that
14:44they're going to have 30% chance of
14:46getting cancer sometime in their life
14:47other than the fact that maybe if
14:48they're at high risk then they should
14:49get a test I would say speaking on
14:51behalf of Angelina Jolie my best friend
14:53that you know anything for a lot of
14:55women when you do have a very strong
14:57family history of breast cancer there's
14:59actually something very importantly
15:01empowering about being able to
15:02proactively do that which is I think the
15:04point of her op-ed in the New York Times
15:06and making that choice whether it's the
15:08right choice or not it's an individual
15:10one but I hear your point that you're
15:11really talking about the fact that at
15:13the end of the day when it comes to the
15:15mass patients in the system it's really
15:17being able to detect
15:19versus predict that's important well
15:21imagine this is another choice for women
15:23to be able to do instead of doing
15:25something so significant as a radical
15:28to be able to just know whether there's
15:30an issue or not right and early mm-hmm
15:32and it's obviously not a problem that
15:33just touches women I mean cancer touches
15:36everybody absolutely I mean I think one
15:38of the most inspiring things as I've
15:41been building a freedom is hearing the
15:43personal stories of not only the
15:44employees but my friends who had all
15:46been touched by cancer in some way shape
15:48or form my grandfather who helped raise
15:50me as well as my dad both have cancer
15:53right now it's it's a story that we hear
15:56all the time from everyone is you know
15:59cancer is traumatic for everyone
16:01involved not just a person that has the
16:03disease and if we can somehow make that
16:06better somehow give them a higher
16:08likelihood of survival we're not just
16:11affecting those patients we're affecting
16:13their loved ones as well yeah this is a
16:15place where technology can do a lot and
16:17I want to hear your guys as thoughts on
16:19what are the trends in this space that
16:22are being brought to bear on this severe
16:23problem you know the major news in the
16:26genomic industry this week which is the
16:28largest genomics company of any kind at
16:30the sequencing layer about two days ago
16:32had a 25% drop in their market cap which
16:35definitely sent waves in the industry
16:37and I was really puzzled because
16:39Illumina effectively is a monopoly in
16:43the space of DNA sequencing and in a lot
16:45of ways rightly so their technology is
16:48very good we have one in-house that we
16:51use for our analysis as well I think
16:54there's a chicken and egg here which is
16:55that they're producing in a sense the
16:57hardware and the software needs to catch
16:59up TJ watson famously talked about how
17:00what there's only be 100 computers in
17:02the world and at a certain age that made
17:04sense but then this offer caught up and
17:07now everybody has computers I don't know
17:09when's gonna get the point where we each
17:10have our genome sequencer at our house
17:12or something like that that might take a
17:13little while but not actually crazy to
17:14imagine by any means just interrupt you
17:16for a second why would we want that is
17:17it just we can personally sequence or
17:19you can imagine like it's a little off
17:22color but you wake up the morning you
17:24use the toilet it sequences everything
17:30that's a lot easier than like taking
17:32blood there's lots of possibilities that
17:34you could imagine and the far future you
17:36know just like it seemed outlandish to
17:38have a computer in your house in your
17:40pocket far less were in your pocket I
17:41mean that just seems ridiculous like
17:43it's like who would need that yeah but
17:46but you know you wouldn't need the
17:48computer in your pocket if you didn't
17:49have something to do with it if you
17:50didn't have the software that's a key
17:51part here and I think what we'll expect
17:53to see more broadly is that prevention
17:55in healthcare used to be like eat better
17:58and exercise and sleep more you know and
18:01there's real limits to what people can
18:02do there but when prevention is catching
18:05diseases early through things like
18:06genomics whether it be cancer or other
18:08areas this is something where I think we
18:10can see medicine really radically
18:13transformed I don't know if I want
18:14expand on the taller now if you've seen
18:21the movie Elysium they have this like
18:23magical diagnostic machine that looks
18:25really cool and people like this
18:27diagnostic machine in future I actually
18:28think the magical diagnostic machine of
18:30the future is your bathroom I'm willing
18:35to buy this I just want to talk about it
18:39and I also told you there haven't been
18:42enough useful clinical applications and
18:44expanded in the market Illumina keep
18:46selling the sequencing machines but if
18:48people aren't building applications on
18:49top of them that isn't good for either
18:51party if you think to the 90s Wintel was
18:54the thing though in a Windows and Intel
18:56together core marketing and that took
18:58both of those companies to new heights
18:59something like that needs to happen in
19:01the genomics world as well there needs
19:02to be a window a Microsoft and in the
19:05genomics world and a lot of people think
19:06that'll be in cancer so clearly the
19:10sequencing layer is just one level we
19:12really need to focus on the application
19:13layer and not just that they're their
19:15layers on top of each other they're very
19:16symbiotic and relationship almost to use
19:19a biological analogy and speaking of
19:21that the broader ecosystem of all these
19:23players like you know big companies
19:24startups what happens when you have a
19:27big player like that because frankly you
19:29only care about a monopoly being
19:30anti-competitive if it's preventing
19:32consumers from benefiting not other
19:34competitors and as long as consumers
19:36benefit why do we care what's the what's
19:38a is there a is that a good or bad thing
19:40I mean one good thing would be is it
19:42the state of network effect the ability
19:43to make a better product for consumers
19:46is obviously very favorable the question
19:48is and what which ways are they good in
19:50which ways there's a poor and you know
19:51there's different aspects of monopolies
19:53here one is a damn monopolies another
19:54one is a sequencing monopoly on the
19:56sequencing side there are several other
19:58companies doing sequencing and
19:59sequencing using different technologies
20:01and different means so you never know
20:03what's gonna pop up I mean the dominant
20:06computer companies from the 70s aren't
20:08necessarily the dominant computer
20:09companies now exactly and so there are
20:11their changes and evolution can be very
20:13positive I mean I want to hear what the
20:15opportunities for startups are obviously
20:16like why you would even try to build a
20:18company in this environment I think what
20:19like Illumina have done well is not
20:23necessarily preventing innovation built
20:26on top of their technologies from
20:27happening I I don't think that's always
20:29the case there have been cases where
20:32companies are trying to be protective
20:33about their technology and how it's used
20:35but by and large we have been able to do
20:37our work without being disturbed and I
20:40think that's very good I think where
20:43there are some worries in my mind is my
20:46PhD working computational biology I've
20:48never touched a machine that's not an
20:51Illumina machine there's this single
20:53technology that a lot of companies are
20:55basing their technologies on top of and
20:58if history tells us anything dominant
21:01players don't remain dominant players
21:03indefinitely if there are so many
21:05companies doing new innovative things on
21:08top of a single platform that may not be
21:10around in ten years what does that look
21:12like yeah although I think that's a
21:14common I mean I'm in the computer
21:15analogy there's lots of software that
21:17was built on top of offering systems
21:18that we don't use right now but then
21:20they get ported to new systems right
21:22people find a way of making things
21:23interoperable they're finding ways of
21:25adapting it it's not so clear-cut but
21:26it's not like the work I mean frankly I
21:28would even argue that some of what
21:29you're describing is really a commodity
21:30layer yeah and it's almost kind of
21:32pointless whose labels on it at the end
21:34of the day and so you kind of only care
21:35about the real value which is the
21:37software applications the data services
21:40and those benefits and interestingly
21:42Illumina has started realizing that as
21:43well and is itself building applications
21:46they bought a company called very nada
21:48which is a ni PT company with ni PT
21:51non-invasive prenatal testing oh great I
21:54yeah you know they have a counter
21:56company called Grail there's a company
21:58called helix so they have started
22:00participating in the applications layer
22:01which is interesting and I'm curious to
22:02see how that works because it's sort of
22:03I mean if you draw this analogy and it's
22:05like Intel building software apps
22:07perhaps they could do but it's unclear
22:09actually any big company making a shift
22:11to a whole new type of business it's
22:12like an on-prem software company
22:14becoming a software as a service company
22:15it's like a hardware company became a
22:17software company a non tech company
22:20becoming a tech company which is
22:21happening everywhere yeah fascinating
22:23wave that it belongs to putting aside
22:26all the issues that the monopolies and
22:27everything else what are some of the
22:29commercial challenges in this space like
22:32I mean why isn't it taken off more
22:34you're totally right there's been a lot
22:35of excitement and interest in genomics
22:38both in the application and see are you
22:40hype yes hype even potentially and we
22:43haven't seen the output of that really
22:44the largest public company on the
22:46application Slayer is this company
22:48called exact Sciences which is about
22:49roughly two billion dollars in market
22:51cap which is small relative to how much
22:53interest has gone into it and I think
22:54it's mainly because reimbursement has
22:57been very difficult that's why you think
22:59yes to the business model aspect of it
23:01not necessarily like any kind of
23:03technological limitation it is a very
23:05business model specific problem because
23:07the way these things work is you need a
23:09provider and actually a lot of it comes
23:11down to the three party system we have
23:13in the US the you know you have the
23:15patient Republic and the patient you
23:18have the provider or the doctor and you
23:21have the pair which is insurance company
23:22and the pay the patient doesn't really
23:25have any say in what gets prescribed to
23:28him or her in order to sell get a test
23:30sold you need to convince the insurance
23:32company to pay for it and then you also
23:34need to convince the provider or the
23:35doctor to prescribe it and that is the
23:37only way this happens the hardest part
23:39in this industry has been convincing the
23:41pairs or the insurance companies to
23:43cover the cost of these tests and maybe
23:47this is a little harsh but I think
23:48they're being a little short-sighted
23:50insurance companies they don't invest in
23:52things that have an ROI window off more
23:55than two to three years and for a lot of
23:57these genomic diagnostic tests you
23:59really realize the value of your
24:01investment in covering one of these
24:03tests in more than three years yeah
24:06this is actually so interesting because
24:07it's the same problem that happens a lot
24:08deductibles you distribute risking your
24:11locked into a single insurance company
24:12then you all see the payout you're all
24:14aligned and incentive because it's
24:16coming to you the long term but when you
24:18have a system where everyone can move
24:19around and switch and change and whatnot
24:21you don't get this incentive for this
24:23long term yes pay off exactly and that's
24:26been a major issue in fact I think in
24:28single-payer systems like the UK or
24:30outside we might see much more adoption
24:32of some of these genomic tests but yeah
24:35that's been one of the main constraints
24:36a is getting insurance companies on
24:38board willing to reimburse there's a
24:40company called Asterix call in mental
24:41health they were making sixty million
24:43dollars in revenue last year but they
24:45were only getting 20% of their tests
24:47reimbursed that's crazy
24:48they were only getting 20% of their
24:50tests paid for despite the fact that
24:51these tests were providing value and so
24:54that that you know that challenge has
24:55been very hard to overcome so how do you
24:57people overcome it like is that gonna
25:00the main pushback there a lot insurance
25:02companies have been giving these genomic
25:04diagnostic companies is that they are
25:06too expensive and they are not accurate
25:08enough and in all honesty I think that's
25:11somewhat true given the technology we
25:13have today we're very good on accuracy
25:14but then usually in the thousands of
25:16dollars and they are useful but at that
25:20price point is it useful enough is a
25:21question right to make that r-la exactly
25:24so what needs to happen is for these
25:26tests to get more accurate and cheaper
25:28and in order for that to happen we need
25:29some of these new startups with these
25:31new technologies you know we've talked
25:32about cancer as an application but what
25:34other applications do you think there
25:36are for genomics clinical or otherwise I
25:39think there's a lot of interesting
25:40opportunities because clearly people
25:42aren't the only organisms on the planet
25:45there are a lot of the interesting
25:46things to do with plants and obviously
25:49you can test plants in a way that you
25:50would never test with people or animals
25:52and you can make better crops and and or
25:54just find better conditions for them and
25:56even livestock to is a very natural
25:59approach that you could understand
26:01better which feeds would help them or
26:03other aspects of improving your herd and
26:05making farmers jobs easier and this is
26:08not like when people talk about eugenics
26:09like cloning and designing it really
26:12people been doing you James for Kathy
26:14whatever for the last 5,000 years
26:20people with their cute little dogs and
26:23their dresses like I'm pretty sure those
26:24dogs have been bred yea they look a
26:25little different from wolves so what are
26:28some of the other interesting trends or
26:29opportunities that you guys think are
26:31ahead just in this space in general
26:33personally one of my favorite stalkers
26:35has proteomics and like all the mass
26:36spec work that's happening there are a
26:38lot of interesting molecules and blood
26:39you know there's proteins there's RNA
26:42there's lipids from you omics is
26:44particularly intriguing because mass
26:45spec is so sensitive and also doesn't
26:48require it and therefore it doesn't
26:50require much sample so and by the way by
26:51mass spec we mean mass spectrometer is
26:53the instrument that actually measures
26:55yeah exactly and the amazing thing about
26:58is that in principle you should be able
27:00to just prick your finger put it on a
27:01piece of paper let it dry mail it in and
27:04mass spec should be able to get
27:06something from that whether that's
27:07useful or not this remains to be seen
27:09but it's a great place also for machine
27:11learning to come in because the signal
27:12that comes out of mass spec does need a
27:14lot of interpretation typically but yet
27:16it's very data rich so in another
27:18opportunity for it to continue learning
27:19and actually benefit yeah absolutely
27:21opportunity for it to learn or to fuse
27:24with other data sets Vijay talked about
27:27novel applications of genetics outside
27:29of humans I think there are also novel
27:31applications of genetics within humans
27:33as well we talked about different types
27:34of applications in non-invasive prenatal
27:36testing and infertility is one major
27:38area in fact that was the first it was
27:40the first to make it big in genetics and
27:42it's gone a little saturated cancer is
27:43another one consume is another one
27:45but there's entirely new verticals of
27:47applications to very new ones our mental
27:49health apparently it's possible to
27:51producing your genetics predict what
27:53type of psychotherapies and drugs work
27:55best for mental health conditions
27:57there's a company called asuric that was
27:58just bought by myriad for half a billion
28:00dollars there's an entirely new
28:02application that's again very nascent in
28:04infectious disease and there's lots and
28:06lots of other new things that we haven't
28:07discovered it's I'm very very interested
28:09in seeing some of those come about yeah
28:11and one of my other obvious favorite
28:13topics is CRISPR and gene editing but
28:16we'll save that for another podcast we
28:17should keep in mind that just from our
28:19knowledge perspective we've just
28:21scratched the surface on how complicated
28:24this entire system is there is the DNA
28:28there is RNA there is protein but then
28:29there's all sorts of variations
28:31within those systems and then we haven't
28:33even begun discussing spatial
28:35orientation of for example the DNA
28:37within the nucleus and how that can
28:39affect you mean like where it's placed
28:40yeah so you know there are DNA and when
28:44it's sort of free-floating within the
28:45nucleus is organized in a very
28:47structured way the 3d conformation so
28:50oftentimes genes that are transcribed
28:53together or actually spatially close
28:55together but they may be on separate
28:57chromosomes the cell knows how to bring
28:59those together so there's maximum
29:01efficiency around that system and we
29:03haven't we don't really even have good
29:05technology to really understand that yet
29:07I think we really have to pause for a
29:09moment and remind ourselves of that that
29:11we're not at the end of the computing
29:13revolution by any means I mean clearly
29:15things are jumping and by leaps and
29:16bounds we talk about deep learning in AI
29:18but in bio and and computer software
29:22we're at the very very beginning you're
29:24right I think we forget that and not to
29:26get all crazy sci-fi but what you just
29:28described reminded me that a cell is
29:30essentially a mini computer in and of
29:32itself and if you think about the human
29:33body as this whole system and a computer
29:35or a set of computers you literally have
29:38like this inception or it's like
29:39computers inside computers inside
29:41computers inside computers it's really
29:42glad working - so I'm really excited to
29:47see what new technologies come about
29:49that's gonna enable us to study things
29:51that we couldn't before and what kind of
29:53new innovations that will create into
29:55health space that we're not able to even
29:57think about today well that was really
30:00interesting and thank you for joining
30:01day six in C bud cast thank you I'm just
30:04really glad you didn't bring up any more
30:05analogies about bathroom and poop
