00:00hi everyone welcome to the a 6nz podcast
00:02today's episode features a 16 Z Partners
00:05Marc Andreessen Chris Dixon and Vijay
00:06Pandey who spent the last year here as a
00:08professor in residence the topic we're
00:10focusing on today is bio and more
00:11particularly the intersection of biology
00:13with computer science we're starting to
00:16see an inflection of many different
00:18things coming together and you know this
00:20is a confluence of many effects on the
00:21on the biology and genomics side as well
00:24as on the computer science side as well
00:26clearly the world has changed
00:28so what's different why now is more bio
00:30innovation possible for more
00:31entrepreneurs than ever before because
00:33of cost much like what AWS did for web
00:35startups or is it about something more
00:37it's not just about cost or capex it's
00:41about doing things that you couldn't do
00:42before we're seeing the ability to do
00:44types of experiments that even with a
00:47huge pile of cash wouldn't necessarily
00:48be all that easy to do and so having
00:51something which in a sense used to be
00:52impossible now become cheap inspires
00:55people to do really new and exciting
00:56things and finally how do we know as it
01:00many innovations that come around before
01:01but take off only at certain times how
01:03do we know that this time is different
01:04you say it's different this time I don't
01:06you know I'm skeptical yeah no that's
01:08very natural I think because things are
01:10always different this time until they're
01:11different okay so that's the
01:13introduction for today's a 6nz podcast
01:15let's hear from mark Chris and Vijay
01:17let's start by you know thinking about
01:19history and then think about what what's
01:21gonna happen in the future so
01:22historically you know for the last 30 or
01:2440 years there have been kind of two
01:25worlds of venture capital and two worlds
01:27of startups high-tech startups there's
01:29been the computer science or IT world
01:31which is well known and then there has
01:33been the life sciences world and the
01:35life sciences world you know has been
01:37characterized by buyout you know terms
01:39like biotech new drug companies and then
01:43also a new medical device companies and
01:44so you know technology stance and all
01:46kinds of new you know all kinds of new
01:47implantable medical devices have come
01:49out of that historically these have been
01:51two very different worlds very different
01:54founder profiles very different in fact
01:57there's been very little intermingling
01:58between the two worlds historically
02:01that's sort of the last 20 or 30 years
02:03do we think that's how things will
02:05continue do we think the things are
02:06changing and if they're changing why are
02:08they changing yeah we're seeing
02:09something that is really a pretty seen
02:11hearing change because it's
02:12two worlds are really not separate
02:14anymore you know you think about
02:15actually students at Stanford some huge
02:18fraction of them I think like 75% or
02:21more take some sort computer science
02:23class some programming computer science
02:24class so we have a new breed of
02:26biologists and chemists and doctors who
02:29not just are familiar with computers but
02:31actually they know how to program they
02:32know the details and so these two camps
02:34are not necessarily separate anymore
02:36we're starting to see something where it
02:37is a combination the two which is
02:39actually an interesting challenge
02:40because it sort of requires expertise in
02:43both of these areas which themselves are
02:44East quite deep and how recent do we
02:47think this this has started to change
02:48yeah I think it's it's like anything if
02:50some of these things where it it grows
02:53slowly and then sort of hits an
02:55inflection point and I think just over
02:56the last year or so we're starting to
02:58see an inflection of many different
03:00things coming together and you know this
03:02is a confluence of many effects on the
03:03on the biology and genomics side as well
03:06as on the computer science side as well
03:08take us through your view of like what
03:10what is a modern bio startup or a modern
03:12converge bio SES startup like what what
03:14are the characteristics of that new kind
03:17of startup yeah there's actually several
03:19things that we're seeing and I think
03:20actually these startups have seen the
03:23success on the software side and are
03:24starting to borrow a lot of those ideas
03:26so one key thing is that they can move
03:28very quickly and move quickly with a
03:29small amount of initial capital and this
03:32is something that's really quite new it
03:33used to be that when you start a bio
03:35startup you'd have to put a lot of money
03:37into building up a lab and building up a
03:39large team so to run that lab and that's
03:41something that's really radically
03:42different the second thing is there's
03:44just a lot of information and so a
03:45that's an opportunity but it's also a
03:47challenge how do you handle all that
03:48information so a lot of ways it's
03:51similar to what happened to
03:52semiconductors and I think around 1980
03:54or so when you add the advent of
03:56fabulous semiconductors which meant that
03:58instead of a semiconductor company
04:00having to fabricate their own silicon
04:04they could outsource it to places like
04:05Taiwan and then just a very small group
04:09with of engineers with little capital
04:11could create an innovative new
04:13semiconductor and that created a an
04:15explosion of new startups and then we
04:16saw something similar with Internet
04:18start-ups in the you know the 90s you
04:20had to spend raise tens of millions of
04:21dollars set up your own data center you
04:24kind of back-end systems and then you
04:26had things like AWS and you know
04:29famously you know you have Instagram
04:31with a couple you know 15 people and
04:33billions of users and whatsapp and etc
04:36it seems like we're seeing some of that
04:38happening now and in biology I think
04:40it's exactly right and it's interesting
04:42there's two ways this is having an
04:44impact one is the obvious that people
04:45are looking on the biology side and
04:47seeing that one can do things this way
04:50and so in a sense you know these young
04:53biologists are on Instagram they've seen
04:56how this thing has worked and very much
04:58inspired by it but the second one is
05:00that there's this great appeal to move
05:02very quickly and so just like you can
05:05move quickly with AWS nobody really
05:06wants to build up a lab and do all that
05:08you want to try your idea very quickly
05:09and see how it goes but then the final
05:12thing I think which is to me the most
05:13important one is that's not just about
05:15cost or capex it's about doing things
05:18that you couldn't do before we're seeing
05:20the ability to do types of experiments
05:22that even with a huge pile of cash
05:24wouldn't necessarily be all that easy to
05:26do and so having something which in a
05:29sense used to be impossible now become
05:31cheap inspires people to do really new
05:33and exciting things so let's yeah let's
05:35talk about some categories let's talk
05:36about some examples say you're at a
05:38dinner party or perhaps on a podcast and
05:40you want to go through let's maybe go
05:42through three or four examples of
05:44categories where we think that this is
05:45happening yeah that's a great question
05:47and I think there's a couple different
05:49categories that we're seeing so one
05:51category that is I think a natural one
05:54because of all the devices and the
05:57social networks have been built as an
05:59area of digital therapeutics and you
06:02know what's intriguing here is that
06:03there are certain cases where it's very
06:04natural that the current medical
06:07approach is really the natural approach
06:09you know if you are in a car accident
06:11and you're bleeding you know there's
06:13certain surgery that you do or if you
06:15have bacterial infection you want to
06:17take an antibiotic but similarly there's
06:19actually some things that I think it's
06:21natural to question whether a drug is
06:22really the natural approach areas such
06:24as in depression or sleeping issues or
06:28smoking cessation in these us more
06:31lifestyle issues which are starting to
06:33become real dominant issues in our
06:34culture today it's not clear that a drug
06:36really is the best approach
06:38and what's intriguing is that maybe now
06:40that with the emergence of digital
06:41therapeutics with the fact that we all
06:43have phones on us all the time that can
06:46sort of detect and understand what's
06:48going on with us combined with social
06:50taking that data and connecting us to
06:52other people allows for the
06:53opportunities and many interesting areas
06:55in these places where maybe drugs would
06:57not work as well mm-hmm so this was big
06:59into the the term a little bit so
07:01digital therapeutic yeah so therapeutic
07:03meaning that it is a solution to a
07:05medical problem it's some level of
07:07keyword or treatment for a problem
07:09digital meaning it's not a drug it's not
07:12an implanted device it's something else
07:13so what would be an example of that
07:15where you feel like there would be
07:17evidence you know the question I think
07:18people would have is okay like literally
07:20you're gonna have an app that can solve
07:22you know diabetes or solve depression
07:24like is that is that is that possible
07:25like what would be an example of where
07:28this might happen and where there would
07:29actually be you know potentially
07:30clinical evidence or reasons to believe
07:32that it might actually work
07:33I think that's the real question because
07:35I think it's very natural for someone to
07:37be skeptical in this area and you know
07:39how could just an app or a social
07:42networking actually have this impact and
07:43actually when I talk to friends of mine
07:45who actually are clinicians and I talk
07:47about some of the work that my lab or my
07:50companies have been involved in drug
07:51design they will often like put their
07:54arm around my shoulder in a very
07:55friendly way and say you know it's not
07:57that we necessarily need more drugs I
07:59wish people would just listen to what I
08:00told them to do if they would just eat
08:02better and sleep and take and exercise
08:04and take care of themselves and it
08:06actually is amazing how much that would
08:07actually have a huge impact on human
08:09health and diabetes is a great example
08:10of that it's you could take a drug to
08:13help your diabetes or you could control
08:15what you eat you could do exercise you
08:17could do all these and this is type 2
08:18diabetes where type C a little bit
08:21larger behaviorally determined to you
08:22yeah that's an important distinction
08:23yeah absolutely right type 2 diabetes
08:25and and so towards that end the problem
08:27the reason why the pill is appealing is
08:29that it's actually hard to actually do
08:31all the things that you're supposed to
08:32do it's not bad anyone wants to get type
08:352 diabetes and they're actively trying
08:37to do things but in these lifestyle
08:39areas that's where an app could MIT and
08:41a network could make a huge difference
08:42and this is what we mean by digital
08:44therapeutics something where it can take
08:46the lifestyle issues that need to be
08:48done and really enforce and help you
08:51along the way towards complying towards
08:52those ends and in fact we have a
08:54portfolio company that does this does
08:55this for diabetes called Omata what
08:58level of what level of evidence do you
08:59believe Omata has like what level of
09:00evidence can Amato present either to use
09:02you know to users or transference
09:04companies or to employers that it can
09:05actually move the needle on something
09:06like diabetes yeah you know it's natural
09:08especially you know for me coming to a
09:10science background that science is
09:11extremely evidence-based and so we want
09:13to see actually does this work and so
09:14just like you could go to the evidence
09:18to see how well a drug works you can
09:19actually compare a placebo in this case
09:22maybe just talking broadly about
09:24diabetes to a to the leading drug to the
09:28digital therapeutic and compare the the
09:30effect in terms of how well the patients
09:32doing in response to diabetes and I
09:34believe actual modest have done this and
09:36and and in this case one can show
09:38actually that that digital therapeutic
09:40is not just comparable to the drug but
09:42and and generally can exceed the
09:43capabilities of the drug yeah so one of
09:46the things you mentioned is we talked
09:47about digital therapeutics you talked
09:48about the role of social networks may be
09:50big I don't think people really
09:51necessarily we wrap their head around
09:52this social network sort of steel view
09:54to something primarily that are you know
09:55most people I think now think social
09:56networks are fun and they're enjoyable
09:58on Facebook and on Twitter and Instagram
09:59and are having a good time you're sort
10:02of sort of hinting at role for social
10:04networks in actually improving people's
10:05health can you talk about like what what
10:07does that mean yeah it's a great
10:08question I think we think about social
10:11networks in terms of communication but a
10:13lot of I think what it is to handle
10:15these challenging issues is to be able
10:18to have the inspiration and the you know
10:22you could pose it as inspiration which
10:23is a positive thing or sort of as a
10:25carrot or stick as someone sort of on
10:27you all the time to make sure you're
10:28doing this so actually and that exercise
10:31better diet less smoking less drinking
10:34all these things I think are in these
10:35categories right and you know having
10:37your friends or people that are a part
10:40of a team associated with the
10:41therapeutic they're sort of to have your
10:43back that makes a huge difference I
10:45think it's just the way human nature
10:47works another thing I find fascinating
10:48is at least my outside interpretation of
10:50the field of medicine as medicine is
10:51very comfortable with there is a problem
10:53and we are now going to fix it which has
10:56in there's been you know tremendous
10:57progress across a broad range of medical
10:59issues over the decades because of that
11:01it seems like more and more if you look
11:05term healthcare spending in the US or if
11:07you think about the health problems that
11:08we all are likely to have especially as
11:11we get older it seems like more and more
11:13they're not just things that happen to
11:15you there are things that happen as a
11:16consequence of things you do as we've
11:18been discussing so you know you don't
11:20exercise you don't eat right you don't
11:22do this you don't get enough sleep and
11:23things will go wrong and and as we solve
11:26more and more of the things that can be
11:27solved with a pill the things that can't
11:30just be solved with the pill that are
11:32kind of lifestyle determined become a
11:33bigger and bigger percentage and I've
11:34seen estimates that you know as much as
11:35something like 75% of long-term health
11:38spending in the US is going to be
11:39correlated to issues that are caused at
11:41least in part by by by behavior and so
11:44it seems like you know sort of
11:46paradoxically it seems like the next big
11:47set of advances in medicine might be
11:49really helping people deal with with
11:51behavioral things and at least my
11:54interpretation is in the field of
11:55medicine and these are viewed as fuzzier
11:56issues or at least harder issues to get
11:59yeah it feels like they'll be rising
12:00much more to the forefront I think so
12:01and that's actually what's that's both
12:03the interesting opportunity and the
12:04Challenger because if you're bleeding we
12:06know what to do or if you have a
12:09bacterial infection we know what to do
12:10in here is trying to take care of these
12:12issues before they become more serious
12:14in the context of diabetes there is this
12:17medical class of being pre-diabetic
12:19which is something of a concern or maybe
12:21you have high cholesterol or there's
12:23tons of markers that you may get early
12:26on to say that there will be a problem
12:28right and what's intriguing is that in
12:30the case of let's say high cholesterol
12:32if you don't do something about it in
12:34your 40s and your 50s the you know
12:37medical implications are quite
12:38significant right but with all these
12:40things which is intriguing is that you
12:41actually can get also get feedback to
12:43see if it's working right you get the
12:45initial indication that there's a
12:46problem instead of taking a drug for one
12:48of these things it would be intriguing
12:50to take to use one of these digital
12:52therapeutics and it's not just something
12:54where you sort of roll the dice and hope
12:55you can actually see the progress that's
12:58digital therapeutics that's one category
13:00or a sort of example of this new
13:01generation of bio companies what would
13:03be another another category that you
13:04think is interesting yeah a second
13:05category that's interesting is sort of
13:07loot so what we're talking about earlier
13:09with the building up of the analog fabs
13:13for semiconductor or for cloud computing
13:15and this is an area that you know we've
13:17been calling cloud biology and this is
13:20not running calculations on AWS that
13:23involve biology these are biology
13:25experiments that are done in a cloud
13:28like way and what makes a cloud like is
13:30instead of one company owning one lab
13:32much like it used to be that one company
13:34owned one data center now there's a
13:36shared lab that's shared by many many
13:38companies and you might run one
13:40experiment on Tuesday in a radically
13:41different experiment on Thursday but the
13:44key thing is that like cloud computing
13:45you pay for what you use along the way
13:47and so you don't have to pay for the
13:48infrastructure but what's even more
13:50exciting is that you can scale up and
13:51scale down the elasticity of AWS is
13:54extremely appealing the ability to run a
13:56calculation on ten thousand servers for
13:58the next twenty minutes and then go back
13:59to three depending on load that's
14:01something that you even if you built up
14:04a data center you maybe don't even have
14:0510,000 servers in your data center you
14:07maybe you have a hundred to average
14:08handle the average load similarly here
14:11the idea that you can do experiments at
14:13scale and elasticity is really exciting
14:15it's both efficient but also in many
14:17cases you might just never have the lab
14:19that could do these things
14:21it also can dramatically improve
14:22reproducibility right that's a really
14:24great point is that the reproducibility
14:26is something that I think is really
14:27underestimated it comes up all the time
14:30that a large fraction of biology
14:32experiments are just irreproducible
14:34and because numbers I've heard of
14:36shockingly high like yes 30 percent yeah
14:39extremely high at times yeah yeah so the
14:47problems that you know some experiment
14:49gets done and gets published and
14:50actually doing something experiment with
14:52mice let's say and you're giving drugs
14:54and yeah observing behavior and you
14:56publish your statistics and it's done
14:58all very carefully in principle and then
15:01actually second group goes to repeat the
15:03experiment right and they cannot be able
15:05to repeat it right and actually there is
15:06even a journal that is whose sole goal
15:08it is to repeat hyper
15:10file experiments because those are the
15:12most important to know whether they are
15:13reproducible or not right and if you
15:15think about how biology is done there's
15:16a lot of challenges and it's you know
15:18done by really brilliant people but
15:20there's just the challenges that a lot
15:22of the process involves a lot of human
15:25labor a lot of people working their ass
15:27off at 1:00 a.m. to get this thing done
15:29to be able to really push science
15:31forward and this isn't really nationally
15:33the conditions of all this doing this by
15:34hand under difficult circumstances is
15:37the best thing for reproducibility or
15:39for for for all these issues that we're
15:41talking let's pause for a second so I
15:42think if you describe this the average
15:44person in the state and if you said a
15:45research group got you know two million
15:47dollars in funding and did an experiment
15:48and published experiment and it can't be
15:50reproduced most people would say oh that
15:51sounds like fraud yeah like that sounds
15:53like somebody was being deliberately
15:54deceptive you're saying that's not
15:56typically what you're seeing it says
15:57okay the case it's something else yeah
15:59and actually the thing about fraud and
16:01science is that the reputation is
16:02everything in science and so if one's
16:05experiments are reproducible that that
16:09alone you know is the type of thing
16:11that's extremely threatening to your
16:12career and so there's there's all this
16:15psychological and cultural reasons to
16:17avoid this but yet even still this is
16:20still a challenge and I think that's not
16:21speaks to not that people don't know
16:24what they're doing or that that there
16:26have other motivations it's just it's a
16:28huge challenge when you're and you have
16:29this sort of awkward thing situation
16:31where you have these very you know
16:33brilliant scientists who are doing their
16:36scientific work and then also you know
16:38administering a test on rats or mice
16:42which is a very different activity yeah
16:43that's right I'm doing it under pressure
16:45I'm you know doing it you know I you
16:50know I don't know all sorts of it sounds
16:52like a different set of skills in the
16:55morning three and more you think about
16:56like just the Industrial Revolution is
16:58maybe not a bad analogy here which is
17:00that when everything is manufactured by
17:02hand right that they'll be greater error
17:04rates and harder consistency machines
17:06just are better at consistency example
17:09here is emerald tara pew --tx which is
17:10very much like this too and very code
17:12driven reproducibility is built-in it's
17:14reproducible the way come code is
17:16reproducible you just rerun it and you
17:19know even in terms of where the robot is
17:20broken these things run they run
17:22Diagnostics all the time so
17:23even know about those types of things so
17:25people who have been in biology for a
17:26long time would probably say you know
17:27there they might say in response this is
17:29but there have been these things called
17:30contract research organizations for a
17:32long time you have been able to
17:34yeah biology experiments for a long time
17:35I think this is really fundamentally
17:37different you know you could outsource
17:39calculations before AWS as well but the
17:42difference between running calculations
17:44on those older systems in AWS was the
17:46sort of essentially lack of friction
17:48that you have with a dia base for
17:49spinning things up and down see arrows
17:52are great and they can they serve a
17:53really important purpose right now
17:54especially in the whole bio in pharma
17:56industry but the amount of friction
17:58between a CRO and what you can get with
18:00cloud biology is dramatic I mean you're
18:02literally as I understand it you're on
18:03the phone with them you're training them
18:05on procedures yes you're doing what you
18:07do with the freelance labor force yes
18:10and as opposed to versus law biology or
18:13literally writing code you're literally
18:14writing code right Yeah right exactly
18:15okay okay good great examples let's do a
18:18third yeah so the third one I think
18:20that's really something that is also
18:22dramatically changing is in the area of
18:25what I would call computational medicine
18:26doctors right now are sort of flooded
18:28with data whether we're talking about
18:30radiology genomics all a series of tests
18:33in each of these cases it's just a huge
18:36amount of information that's really
18:37getting to the point where it's beyond
18:39what an individual human being can
18:40incorporate and understand even just on
18:43a single test the resolution now that
18:45you can do on some of these imaging is
18:47so great that it's starting to even
18:49surpass which you could look at very
18:50carefully with the human eye it's all
18:53these changes challenge of handling
18:54handling the flood and this connects to
18:56all these beautiful current trends on
18:58the computer side of data science and
19:01machine learning and computers have
19:03become extremely good at handling huge
19:05amount of data and for finding small
19:08patterns within the noise and it's
19:11getting to the point where computers are
19:12exponentially increasing in their
19:13capabilities here and people's
19:15capabilities are have been flat over the
19:17last few thousand years
19:18I think they excite anything here is
19:20that it's not something where the
19:22computer is going to be the doctor or
19:23put the doctor out of business just like
19:25we use word processors rather than sort
19:27of writing by hand or typing they're
19:29huge aid to us to make us more
19:31I think these tools will make the doctor
19:33can Sibley more productive because you
19:34can just do so much more
19:36what would you say like if you ask some
19:37people in biology and you if you make
19:42these claims you're making now to some
19:43people in biology they'll say oh we
19:45heard this before we heard this in the
19:4680s we heard the computers we're gonna
19:49change everything and yeah we use
19:51computers and they're nice and emails
19:52great but fundamentally it hasn't
19:54transformed our industry you know what
19:58what you say it's different this time I
20:00don't you know I'm skeptical ya know
20:02that's very natural I think because
20:03things are always different this time
20:04until they're different and so and we've
20:08seen this in other areas I think you
20:10look at the success of big companies
20:13that are driven by data like Google and
20:15Facebook and they've been very
20:17successful because of what they can do
20:18on the computer side and they're
20:20integrating a huge amount of data and
20:23and so I think that's part of it but
20:26actually maybe what's special now is a
20:28confluence of that with many other
20:30trends there's the fact that Moore's law
20:31has been chugging away and continuing to
20:33make a huge impact and game to the point
20:35now where we can do computations
20:37routinely that we just essentially used
20:39to be impossible but you know this is a
20:41special time in other areas genomics is
20:43making it such that we can sequence your
20:45human genome or your biome or a cancer
20:48tumor and do that cheaply and routinely
20:50and then finally I think and this is
20:52relevant on the this digital therapeutic
20:55side too the ability to have all these
20:57sensors whether we're the sensors in
20:59your phone or sensors that are sort of
21:01being involved in the experiments in
21:02cloud biology the cost of sensors going
21:04to zero connected with mobile is also a
21:07unique opportunity right now so yeah I
21:09think you know we can only prove that
21:11this is something that has happened
21:13until after two efforts happen but there
21:15is something very special about what's
21:16going on right now and we see all these
21:18parts coming together yeah some question
21:21about genomics the role of genomics and
21:23the genome so you know I think if we
21:24were sitting here I don't know 15 years
21:26ago or 12 years ago there was a huge
21:28excitement around decoding the human
21:29genome sequencing the human genome and
21:31there were these huge projects the
21:32government there's this race between the
21:33government and craig Venter to do the
21:35first sequence the first human genome
21:36and it was this this giant project and
21:39you know there were very well respected
21:40experts in the field who at the time I
21:43think predicted you know a revolution
21:44and cures of you know for cancer and all
21:47you know that was sort of quickly follow
21:48I think if we were sitting here five
21:50years ago or three years ago we would
21:52probably say that that had been a bust
21:53and that while we decoded the genome and
21:56found out that it actually didn't give
21:57us many new options in terms of
21:59treatments or drugs where do you think
22:01things stand with genomics today yeah
22:03you know I think this probably follows
22:05the very classic Gartner hype curve in
22:07that there's for any given technology
22:09there's gonna be a point where people
22:11are excited beyond maybe where the
22:13technology is and then the back part of
22:15the hype curve is it's not as exciting
22:17as the peak but is where things become
22:19real and so I think we're way past this
22:22or the huge amount of original
22:24enthusiasm and we're starting to see it
22:26become real there was a huge amount of
22:28money that went into the original human
22:29genome project and that makes it
22:31possible to have cheap genomic
22:33sequencing now mmm-hmm and so it wasn't
22:36that one perspective like it cost
22:37billions 15 years ago and it's now
22:41$1,000 now $40 $40 to sequences and but
22:45if we didn't get huge results out of the
22:48first one that was very expensive why do
22:50we think we'll get better results out of
22:51a million that are much cheaper yeah
22:53that's a great question and I think I
22:55think it was a misunderstanding for
22:56maybe how the genome could be best used
22:58I think well the way people are viewing
23:01it today is that the genome is useful to
23:03tell us about how we're all different
23:04from each other and this has such
23:08wide-ranging implications it could tell
23:09us which drug to give you versus someone
23:13else that's radically different it could
23:15even be in areas of cancer cancer tumors
23:18are constantly changing and knowing what
23:20drug to give is extremely difficult
23:21question and the ability to sequence a
23:23cancer tumor and use that information to
23:26tell you which drug to take is extremely
23:28dramatic one one one example I loved is
23:31it I heard recently from an entrepreneur
23:34is that there are you know better than I
23:35with hundreds of seemingly effective
23:39treatments for cancer but the big
23:42problem is knowing which one to use in
23:43which cases and there's been a lot of
23:46companies like foundation medicine in in
23:51building technologies to do that
23:53matching and it turns out because of the
23:55drop in these in the costs for
23:57sequencing and all of the kind of
24:00in this process that it now becomes in
24:03many ways a software problem so in some
24:06ways we have 200 cures for cancer we
24:08don't know which ones to use and it's
24:09figuring that out as a software problems
24:11that right that's exactly right and I
24:12think this is very clear on the cancer
24:14side where there's a huge number of
24:16drugs and it's never gonna be that the
24:19cure to cancer looks like a single drug
24:20it could be sort of like with AIDS like
24:23it's a cocktail or yeah or a mixture of
24:25drugs but even at least an AIDS it's a
24:26single virus cancer is really an
24:29umbrella term for many many many
24:31different so having a drug to cure
24:32cancer is like a drug to cure disease
24:34yeah you know so it's not a mistake and
24:36kind of a misunderstanding this
24:38understanding really what how a
24:40heterogeneous cancer is the step one is
24:43is is identifying the type of cancer
24:45yeah and and and find the cancer and
24:48then figuring out the initial drug but
24:49the part that actually is a surprise is
24:52that initial drugs can fail to work and
24:54that's scary but actually the upside is
24:57that drugs that didn't work at early
24:58stages will work at later stages and so
25:00now the challenge is how to figure out
25:01which drug to give at what stage and
25:03what time there's that would be
25:04impossible without genomics and without
25:07understanding where that tumor is and so
25:09it's intriguing that you know there's so
25:11much of us that are different and that
25:13those differences change in time whether
25:15we're talking about our cancer tumors
25:17changing or even you know I think you
25:20know we forget that so much of our
25:21bodies is actually just by cells is not
25:24human cells but bacterial cells in our
25:27gut and the microbiome in our gut is
25:29also constantly changing and so being
25:32able to understand those changes at a
25:33molecular level now is possible which is
25:35extremely sci-fi kind of stuff you know
25:38and now has become quite routine but the
25:41routine part is getting the data the not
25:43routine part is what do we do with the
25:45day and that becomes I mean what you
25:47have in these things is you have sort of
25:48this this dramatic change when it goes
25:50from a thousand dollars to test your
25:52stomach about unig stomach bacteria to
25:55close to zero dollars right then you can
25:57do it all the time you can see newest
25:59Limon etre it you just get a whole new
26:00set of things activities you can do
26:02right yeah that's exactly right for all
26:04these kind of magic thresholds in the
26:05same way we have with in computing we've
26:07seen this happen with storage and
26:08processing like storage is effectively
26:10free and that means you can store all
26:12your photos online and you can do you
26:14all these other wonderful things to do
26:15on the internet and social networks etc
26:18you know it's this a big there's a big
26:20big big difference between just a free
26:22and $100 in this case right yeah yeah
26:24exactly yeah so let's talk about let's
26:27talk about economics the implications of
26:29what Chris is talking about in terms of
26:31how startups in biology work so I think
26:34if you if you say you know bio startup
26:36or life sciences startup to an
26:37experienced VC part of the look a horror
26:39that you'll get is the idea of major
26:41regulatory hurdles and you know we've
26:44we've talked in the past about there's
26:45you know the IT side of startups the
26:48computer science side of startups is
26:49driven by Moore's law which is this you
26:51know incredibly fast exponential decline
26:53in the cost of chips we you know there's
26:57this concept on the bio side of
26:58something called eros law which is
27:00literally Moore's law reversed instead
27:02of more it's room rooms law which is and
27:05I don't know exactly what it is but the
27:06the the price of getting FDA approval
27:08literally the ticket price the cash
27:10required to get FDA approval for a new
27:12drug or new medical device have
27:13skyrocketed over time
27:15yeah and for a lot of new drugs now the
27:16price I think is in the billions of
27:17dollars yes is you know way beyond what
27:19a what a start-up people call it
27:21earu moi which is Moore's law backwards
27:23because it's followed the opposite
27:24pattern or Moore's law witches yeah
27:26right and so by bio startups in in
27:28pharma and and medical devices now have
27:30become terrifying to investors from from
27:32a capital requirement standpoint and in
27:34fact a lot of new biotech startups get
27:35bought by big drug companies early now
27:37because it's only the big the big pharma
27:39companies that can really for they spend
27:40hundreds of millions of dollars from VCS
27:42then go public to try to raise more
27:43money try to get through the FDA process
27:45right and then may or may not get that
27:46money yes right and it's just a it's
27:49just a really really expensive endeavor
27:50yeah so these new these new you know you
27:52went through the three categories of you
27:54know these through sort of digital
27:55therapeutics went through clawed biology
27:56and then we went through new kinds of
27:58computational medicine as are three
27:59examples so are these new they're these
28:01new kinds of bio startups are these also
28:02subject to Iran's law yeah that's a
28:04great question and you know what is the
28:08sort of interesting aspect of these is
28:10that they don't have the same profile in
28:11terms of FDA regulation digital
28:14therapeutics is not coming up with small
28:16molecule drug and so therefore it
28:18wouldn't be something that would go
28:19through the typical phase clinical
28:21trials and in each of these cases it it
28:25doesn't have the same type of exposure
28:27and so that actually has the hope to be
28:29fundamentally different combined with
28:31the fact that it has its own Moore's Law
28:33it has Moore's law for computer for
28:34genomics and so while humans law is
28:36exponentially increasing cost for drugs
28:39the fact that what drives all of these
28:41areas are essentially software and
28:43computation and/or genomics and those of
28:45those costs are exponentially decreasing
28:47we would expect that there should be a
28:48radically different behavior so these
28:50new these new startups have the
28:51potential they have the kind of economic
28:53profile and the tenth that the kind of
28:55financing needs of a software startup as
28:57compared to a pharma startup yeah you
28:59know if you think about it these new
29:00startups you know remind me a lot of
29:02software starts in 2005 when we're
29:04starting to see the cloud computing
29:07other things start to realize that's
29:09sort of what we're starting to see now
29:10and because they have software at sort
29:12of their heart either literally or in
29:15terms of how they think about things
29:16that they're organizing themselves in a
29:18sort of cloud like biology way or so on
29:20this would be very much on the Moore's
29:22Law curve of things and in a sense you
29:24could use this to differentiate
29:25traditional biotech from this new crop
29:27of companies that traditional biotech is
29:29governed by Ohm's law and these are
29:31governed much more by Moore's law if you
29:34started at com in 1999 it was probably
29:36an initial out like an Internet company
29:38nice and that guy it's probably an
29:39initial outlay of twenty million dollars
29:40to get going and you had to write big
29:42checks to these big companies like Sun
29:43and EMC and Cisco and Oracle did just
29:45get a website up and running you know
29:48kind of as you allude to by 2005 and
29:50certainly by 2008-2009 cloud computing
29:52and AWS and open source and all these
29:54trends in computer science had evolved
29:57to the point where you have this boom of
29:59angel funding and seed funding starting
30:01new companies not for twenty million
30:03dollars but as an example Facebook got
30:05started on $500,000 and then you had
30:08other examples you have other examples
30:09of very successful companies since then
30:11where the initial seed funding has been
30:12$100,000 or $50,000 in an extreme case
30:15you know now new internet startups get
30:17started and it's you know three is you
30:19know three kids in an apartment you know
30:20in their entire capex budget is their
30:22laptops yeah and whatever supply of
30:24ramen noodles they need for nine months
30:25right and it's like you know it's like
30:26you can like start a new Internet
30:28company and like run an experiment with
30:29a new product for literally hundreds of
30:31dollars and you go on AWS and light up
30:34and go global and see if it works and
30:35like you're out you can put the whole
30:37thing in your credit card so do you
30:39think we'll buy oh you think this new
30:41of computer science driven by Oh we'll
30:42get all the way there in other words is
30:44it is it possible that we will see an
30:46explosion of experimental bio startup
30:48yeah well we're already starting to see
30:49that I think numerous companies that are
30:52coming out of places like Stanford or
30:54all over the place where it's graduate
30:56students who have grown up with the
30:58biology and growing up with computer
30:59science they've got all that the
31:01intellectual tools they need and they
31:03see a device and they see emerald or
31:05they see Klaus Sarah and their brains
31:08are just clicking that basically half a
31:10million to a million dollars is all they
31:11really need to get the job done and that
31:13could get you something through
31:15preclinical which is what Mouse studies
31:17are such I could you could have
31:19something go into phase one or phase two
31:20with that initial seed funding yeah and
31:23that's dramatically different normally
31:24it would be like you know Series D or
31:26ask that or post IPO before you even get
31:28into sort of later stage clinical trials
31:30right right so the consequence could we
31:32could we have the potential to see an
31:33explosion yes sort of seed stage bio
31:35companies running experiments in all
31:37kinds of different areas the failure
31:39rate might be high but it doesn't matter
31:40because you've got so many more
31:41experiments you get a lot more successes
31:43coming out of it yeah that's exactly
31:44right I think that's what makes software
31:45so appealing is that you can fail very
31:47quickly and understand and for a low
31:50cost and here what I think we'll see
31:51exactly the same thing okay great
31:53let's maybe let's spend the last few
31:55minutes talking about you so maybe if
31:58you could give us a thumbnail sketch of
31:59your of your background yeah
32:01and then I've got a few questions about
32:03the ventures that you've been involved
32:04in yeah sure so I've spent the last 15
32:07years at Stanford but before then and
32:10and during then I've been an
32:11entrepreneur so actually you know my
32:13first involvement with computers was
32:15actually I we moved from Long Island to
32:20suburb of Washington DC and that summer
32:23my parents did two things which was
32:24dangerous we moved to a new place where
32:25I had no friends and I spent the summer
32:27with a computer from there from there so
32:31I was in my first startup 15 and this
32:35was a naughty dog a software a computer
32:37game company and that was actually a lot
32:41of fun and exciting to be involved at an
32:42early stage and you know some of my
32:45initial love programming is start to
32:46become lucrative which is also nice
32:47especially it doesn't take that much
32:49money to make a 15 year old happy
32:51and syrups can do such things so then I
32:54went to college I studied science mainly
32:57physics but with a biology a sort of
33:00twist to it and on the academic side my
33:03work has had sort of interfaces with
33:06biology and chemistry and computer
33:07science and so on and at Stanford I
33:10believe there are about PhDs in physics
33:12yeah my PhD is in physics but actually
33:14I'm at Stanford I am in the chemistry
33:17department primarily but also in
33:18structural biology and computer science
33:20and a chair of biophysics so sort of the
33:22whole sort of buffet of all these things
33:24and but you know it was interesting when
33:27I got to Stanford with him I think
33:30literally within weeks various venture
33:33companies asked me to evaluate companies
33:36venture firms ask me about your
33:37companies and this is not an uncommon
33:39thing and it quickly made me introduce
33:43to these companies and I became advisors
33:44to them and that was it's just part of
33:46the fun thing about living here and and
33:48being connected in the ecosystem and did
33:51you realize that before you got her no
33:53actually I did not I was at Berkeley for
33:55four years before then and at that time
33:58the two were actually very different now
33:59I think things are starting to bleed
34:01into there as well yeah but that was
34:03actually really exciting to me because
34:05it was what often happens is that you do
34:07basic research and people ask well
34:10that's great but can you impact disease
34:11you learn about the disease they say
34:13that's great but can you make a drug you
34:15know you make a drug and it's like
34:16that's great but you can you get it into
34:18the hands of patients and at each stage
34:20I felt the sort of push towards being
34:23able to make an impact greater greater
34:24and having these connections with
34:26companies it's a way to have that impact
34:28got it and then now there's two big
34:30projects you've been involved in I'd
34:31love you to describe so one is folding
34:33at home yeah yeah so my group founded
34:36folding at home in October of 2000 so an
34:39hour actually almost up to its 15th
34:41anniversary well so folding home gets
34:43people throughout the world to donate
34:44computer time and I think a lot of
34:46people don't realize this that in in
34:48calculations involved in biology and
34:50chemistry computer time could still be
34:52very much the rate limiting factor that
34:55North to do calculations that we've
34:56needed it would normally take even on
34:58the most powerful supercomputers decades
35:01to hundreds of years to get these things
35:03so that's the first surprise how much
35:05computer powers needing a second
35:06surprise is just how much computer power
35:07is sitting all around us the most
35:10powerful supercomputer in the United
35:12States for science is about at that
35:14twenty petaflop level so that's 20 times
35:1710 to the 15th powers and 15
35:20oscillations per second of operations
35:21per second yeah exactly you know if I
35:23was like if you think a person with a
35:25calculator is one operation for a second
35:27right and you have a billion let's say
35:29ten billion people in the world which
35:31doesn't up over estimate you know it
35:33would still take it would still take ten
35:37to the four so ten thousand worlds right
35:40of people with all those calculators to
35:42do that type of to that type of what a
35:44supercomputer can do yeah yeah but you
35:46know that 20 petaflop s-- is really
35:48quite small and another way think about
35:50is like each GPU puts out right now a
35:53modern GPU easily puts out a teraflop
35:55GPU is a card in your pc that you use a
35:58game playing call of duty yeah yeah it's
36:01a GPU that's putting all those pretty
36:02pictures yeah exactly but it does a ton
36:05of math and so if we just put a million
36:08GPUs together that gets us to a thousand
36:11pedo flops or an exaflop right which is
36:14like this huge holy grail of computing
36:16hey now that's only a million GPUs you
36:18know there's like a billion GPUs on the
36:20planet right so there's just so much
36:22compute power that goes underused you
36:24have 400,000 now I'm full yeah so we
36:26have about 40 petaflop sand as a mixture
36:29of 400,000 computers and a whole bunch
36:31of GPUs and and these are just people
36:33who are sitting at their desk at home or
36:35having fun or experimenting or find this
36:36stuff interesting yes and they download
36:38an app on a PC or under their phone and
36:40it contributes into this kind of global
36:41supercomputer that's exactly right and
36:43we have the algorithms to be able to get
36:44these guys connected it's fun by the way
36:46because you get to see it's like a cool
36:47picture on your screen so what does
36:50folding at home do like what is the
36:51thing that it does yeah so we're trying
36:53to understand the fundamental process
36:54for how proteins is part of the body how
36:57they work and when proteins work well
36:59they do all these great things act as
37:01enzymes they allow biology to happen but
37:04when they when they don't assemble
37:06correctly you get diseases like
37:07Alzheimer's or cancer or several other
37:10things and a lot of these diseases are
37:12very difficult to understand
37:13experimentally through just test tube
37:15experiments and much like you know in
37:18design cars we design bridges we will
37:21simulate them first before going into
37:23the lab and doing something here if
37:25computers could be powerful enough we
37:26could simulate and do things that you
37:28couldn't do in the lab and they make
37:29very specific suggestions and having
37:32done these simulations we can actually
37:33have made predictions for various drugs
37:35and drugs for various different diseases
37:38and with those in mind then it usually
37:41sort of I take off that condemning hat
37:42because we want to push it forward and
37:43then it goes for more into the company
37:45space and that's where we we've been
37:48able to have I think sort of impact in
37:50both areas and so for people listening
37:52if you need it if you want to see pretty
37:54pictures in your screensaver and
37:56contribute to the to solving disease in
37:58the world you download you go to the
38:00folding at home website yeah unfolding
38:02at stanford.edu folding down
38:04stanford.edu okay this was a newer set
38:06at home it was sort of inspired by
38:08settied home yeah very much so and so
38:09deeply understands work with SETI I
38:11think was a great inspiration for many
38:12of us I think the challenge was that he
38:14does search for extraterrestrial life
38:16yeah looking through radio signals from
38:19exactly the challenges that they have it
38:22hasn't yet and can't tell if they've had
38:27better we're also positives right
38:29they've had they've had they've had
38:30things pop up that turn out to be a
38:31microwave oven down the hall and I think
38:35the challenge this is a challenge much
38:36like any serving thing called competing
38:38how do you use like a million processors
38:40you know efficiently do your calculation
38:42so user challenge sort of traditional
38:44computer science challenges of the
38:45modern era okay and then let's wrap up
38:48let's talk about globe of here and I'm
38:50really fascinated by globe of your
38:51string and maybe describe kind of your
38:53role and then just describe with it what
38:54the mission of the company isn't and how
38:56it's going yes so global views a company
38:58involved in infectious disease and areas
39:00related to it and I think are you know
39:02the challenge ahead of us is that there
39:04are many areas infectious disease that
39:06are huge nightmare for us that we don't
39:08have drugs for our therapies for you
39:11know we had this crisis recently about
39:13Ebola and that's something which was a
39:14huge disaster in in Africa but a lot of
39:18scare here too but you know there's
39:20other examples a dengue fever is
39:22something that kills a hundred thousand
39:23people a year and Asia actually my
39:27relatives this is something that's on
39:28their mind the ones that live in India
39:30another area that we've been interested
39:32in globe revere is also the area Shah is
39:34disease Shaka's disease a top-five Cohen
39:36or Latin America but due to global
39:38warming and other and immigration of the
39:40things these insects are moving north
39:41and this is becoming a serious issue NPR
39:45called it the new HIV in America look
39:48and so for all these things these are
39:49infectious diseases which we don't have
39:51existing drugs but our interest has been
39:53to use computational methods to
39:55repurpose drugs to be able to find new
39:58purposes for existing drugs and this is
40:00something where it allows us to move
40:04very quickly and in this sense it's very
40:05much in the spirit of these other
40:06companies we've been talking about its
40:08computational heavy being data-driven
40:10using that that data in intelligent ways
40:13using algorithms and then moving in ways
40:16that can sort of sort of accelerate the
40:19regulatory process so instead of taking
40:2115 years and you know a hundred million
40:23dollars to do the first part we could in
40:26the case of our drugs for a lot of these
40:28areas we could do it in nine months and
40:30also drugs already been shown to be safe
40:32yeah you need to show it's effective or
40:35something something that wasn't
40:36originally designed yeah or tested for
40:39weather exactly I think part of the
40:41surprise here is I think most people's
40:42knee-jerk reaction is well how could
40:44something be useful how clear drug be
40:45useful for two things like isn't it hard
40:47enough to be useful for one thing and I
40:49think the reason why this is such a
40:52surprise is that the brand names given
40:54to drugs masks the chemical names and so
40:57there's many drugs that actually are the
40:59exact same thing but are for use for
41:00different indications like the drug for
41:04in unisom is the same drug in benadryl
41:06and it's actually it's masked by the
41:09fact that it's different chemical names
41:10and this is actually there's many
41:11examples of this and so you know so the
41:14question isn't could a drug serve
41:16multiple purposes that's actually been
41:18empirically shown in many many cases the
41:21question is could you identify what
41:23other what drug can do the purpose you
41:25want and that like a lot of these data
41:27questions is a very natural
41:29computational question and so we're able
41:31to address that and then move very
41:33quickly and I think what I like about
41:36this approach is sort of how computation
41:38really come in and make an impact but
41:40also how we can sort of do our best to
41:43try to speed things through
41:44the normal regulatory environment great
41:46well we're at our time so thank you very
41:48much Vijay and with this will be the
41:49first of many on these very exciting
41:51topics yeah great thank you