The a16z Show - Personal Genomics: Where Are We, Really?

Episode Date: January 6, 2020

This is a turn of the decade (and January-themed) look backward/ look forward into personal genomics, given recent and past retrospective and prospective pieces in the media on the promise, and perils, of the ability to sequence one's DNA: What did it, and does it, mean for personalized medicine, criminal investigations, privacy, and more?General partner Jorge Conde, who has a long history in the space, covers everything from where genealogy databases and large datasets come in to fetal testing, multi-omics, and other themes spanning the past, present, and future of personal genomics in conversation with Sonal Chokshi for episode #18 our news show 16 Minutes, where we cover recent headlines, the a16z way, from our vantage point in tech -- and especially what's hype/ what's real. While we typically cover multiple headlines, this is one of our special deep-dive episodes on a single topic. (You catch up on other such deep dives, on the opioid crisis and other evergreen episodes, at a16z.com/16Minutes). And if you haven't already, be sure to subscribe to the separate feed for "16 Minutes" to continue getting new episodes. image: Petra Fritz / Flickr Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Hi everyone, happy new year. I'm Sonal. As you may know, we launched a new short form news show last year, 16 minutes, where we cover recent news the A6 and Z podcast way, what's hype, what's real, why they matter from a vantage point in tech, and that show has continued in a separate feed for quite some time now. You can subscribe to it if you haven't already in your podcast app by searching for 16 minutes A6 and Z. But I'm also sharing the latest episode here in this show feed, since we sometimes cover not just multiple news items, but a single topic prompted by recent head. headlines like we did on our episodes on e-sports and the opioid crisis. This week, the topic is personal genomics, the promise, the perils, where are we really today,

and where could we be going next? We start with an article by Peter Aldhouse, unquote, 10 years ago, DNA tests were the future of medicine. Now there is social network and a data privacy mess. The article refers to a series of events, everything from companies like 23 and me and the FDA, to some of the headlines we've seen lately around criminals being caught based on their relatives' DNA. There's also a number of companies cited in the article who offer such tests. To be clear, none of the following discussions that should be taken as investment advice,

please see A6.com slash disclosures for important information. So that's a context. And super quick summary. Now, let me introduce our A6 and Z expert, general partner Jorge Condé, who has a long history in this area. Since it's a turn of a decade and the first episode of January, I thought it'd be great for us to do sort of a Janus-themed look-back, look forward, starting with quick reactions on reading the piece. Well, when I read the BuzzFeed piece, which was super interesting, it took me back to a very specific moment in time. And I was living in this world. I was in the personal genomic space. I had just started a startup that was looking to essentially interpret full genome data at scale. If there was something in DNA that could be found to be relevant or

actionable, we were building the technology to detect that. But what I thought was really neat is I'm reading this 10-year retrospective. If I go back to that moment in time, I actually participated in a piece that was in some way a 10-year prospective look on what the future of personal genomics would look like. And this is in the 2008 time frame, more or less, I get an outreach from, of all things, GQ Magazine. They had an author, a guy by the name of Richard Powers, who had just written a book, had won all kinds of awards. He wanted to write about the experience of what it would mean to have his full genome sequenced and essentially revealed to him. And we had started this company Nome with the idea that we would be among the first

to fully sequence individuals and interpret their DNA for them. But what's really interesting is if you almost read that piece as a companion to this backward-looking look, you get the forward-looking look of what the next 10 years in personal genomics would look like. What was it called? It was called The Book of Me.

Oh, fantastic. What a great title. So then what is your take? What's hype? What's real here? When it comes to the promise of personal genomics? The whole complaint of this article is that we were promised one thing. They were supposed to be the future of medicine, but hey, instead we got this big data privacy mess.

So looking 10 years back of what was hype, or at least over-expectation, was that people in general would have a deep curiosity to understand their DNA. You're saying that part is hype? I would think that part is reality. Ah, well, what's really interesting is if you look at several companies named, I think all had at some level an idea

that there would be a large number of people that wanted to very deeply understand any sort of secrets or actionable insights that you could draw from your own genomic information. And while those people definitely exist, I don't think that a large market materialized around those people. In fact, one of the eye-opening things for me when I was starting my company back in 2008, Ancestry.com was primarily selling subscription services for getting into these sort of Ancestry Databases. Yeah, online family trees. Yes. So I remember I downloaded the S1 ancestry.com's subscription revenue with something on the order of $200 million that year. So the question is, do people fundamentally want to understand their DNA in terms of health

risks and the like? Or do people have a fundamental curiosity to know who they are and where they come from? Oh, that's where you're saying the difference between what's the actual market for this kind of, there's a curiosity, but not necessarily a market for DNA around it. Exactly. So people want to understand who they are and where they come from. And if it happens to come from DNA data, great. If it happens to come from looking at ancestry databases, that seems to be a pretty reasonable substitute for getting that insight. So, okay, so you're saying one of the things that's hype is that people may not necessarily want DNA data specifically. What else is hype? I think one of the other things that was potentially hype, certainly at

that time in 2008, 2009, 2010 time frame is that there would be something deeply concrete about DNA that would determine what your potential health risks and therefore what your potential outcomes might look like. You know, sort of it's this idea that DNA is destiny when it comes to your health. Now, that's certainly true in some subset of diseases. The subset of diseases that are known to be monogenic. Right, so single factorial driving it.

Exactly. When there is a mutation in a gene that results in a specific condition, like a sickle cell anemia. Right, which we talked about in our CRISPR episode. But when you start to look at things that are much more complex, much more multifactorial. Like cancer, many other diseases. Cancer, metabolic disorders, you know, pick any number of cardiovascular risks.

There are certainly genetic contributors, but as a lot of experts in the field say, you probably get that same level of information from getting a good family history. Right. So basically, the second hype piece you're saying is that it is not a direct link, a map from, oh, here's your DNA, and then, oh, here's all the diseases you're going to get, not get, et cetera. And here's the precise risk you have. for this disease based on me analyzing your DNA.

So I think that's probably an area where expectations were probably higher than where we were in reality in terms of how actionable is this information for someone that is seeking to manage their health. So that's maybe one of the things where maybe the promise hasn't quite come through yet. And I think another area that is really interesting is a lot of these businesses were conceived as subscription businesses, where I would give someone a DNA kit for Christmas, they would get their genome scan, and then they would engage with that on some regular basis. basis. And I would suspect that the vast majority of people that had those DNA scans done,

oh, so many Christmases ago, probably haven't logged in in a while. So if you had a genome scan done in 2009 and you did another genome scan in 2019, I can almost guarantee you that your ancestral makeup would look different over the course of those 10 years. Simply because of the available data. We just know more. That's right. You, of course, haven't changed who you are, but who an ancestry map tells you you are, has changed. Okay, so that's where maybe things were hyped or not delivered yet or promised and didn't quite come through. Now let's quickly talk about the reality. So where are we today? What is possible right now truly with personal genomics? Well, the first thing I would say is genomics more broadly has delivered a lot over the course of the last 10 years.

In fact, I will say this as an expert, not as an entrepreneur in the genomic space, but as a parent of many children, one of the fascinating things that I saw was the time my wife was pregnant with our oldest child, you still could not get enough of a signal from maternal blood as to whether or not there was sufficient fetal DNA in circulation to determine whether or not there were genetic abnormalities. By the time we had our last child, that was routine, standard of care. The other example is, you know, when the child is actually born, the mandated genetic tests when a child is born by a state, some of the ones you could also opt into, that menu of tests that were available multiplied in the relatively few years between the time when we had

our first child and our last child. So roughly a decade span. So we've seen a lot of advance just in the use of genetic information and the practice of medicine, and that's a remarkable advance forward. So now let's focus on personal genomics specifically. One of the promises of personal genomics, even back in 2009, was predicated on the fact that there would be power in numbers. And in large part, that's why some of the leaders in this space, whether it's 23 and

me or Ancestry.com, that eventually came into this, there's so much value in them amassing a large database. Because in some ways, as you have more samples in a database, you get better reads on who we are, just genealogical. You have a higher resolution map of the world. Now, the risk of having a large aggregated data set, it also becomes attempting target. sometimes for legitimate uses, for investigation, sometimes perhaps for illegitimate uses. Right. This is where privacy concerns come in, exactly.

Very interestingly enough, this privacy question, it sounds very futuristic, but even in 2009, these concerns were very real. If you read the terms and conditions that services had to their credit, they were very explicit that this information, you know, could be used in unintended ways. Oh, in fact, the article even points out that one of the companies had to actually expand their definition of a violent crime in order to cover it in their terms of, and services. And secondly, that some of them are actually moving to opting in to whether you can even be included in that aspect of that database, which is also fascinating. People can actually choose. Well, in Nome, we made a decision early on where we said we're actually not going to

aggregate all of the data. We're not going to centralize it. We did something inverse. What we decided to do was we would sequence an individual and place that sequence that genomic data on an encrypted key that would live in a decentralized network. And the thought was you could keep the queries centralized. So let's say a researcher wanted to understand how many people in a population have this mutation associated with this disease. You would push the queries down to the edges of the network. The analysis would run locally. The result, and only the result would come back, get centralized. Then you'd have an aggregated answer to that question. That's fascinating. Funnily, even though it's a very different example, reminds me of differential

privacy. And that was also something that Apple made a bigger deal about in the last few years, but in fact was based on a paper from Microsoft researchers like a decade ago. It's a fundamental insight they have for how to separate these two things. So it's kind of funny, the synchrony of all that. Yeah, and arguably we were 10 years too early. We came up with it. That's always about timing. The other thing that when we thought through these questions of privacy, a lot of these were perceived risks. We didn't know, but we wrote a lot of risk factors out to getting yourself sequenced. and among them we had things that sound fantastical like if someone had an entire readout of your genome,

they could essentially synthesize your genome and then plant your DNA at a crime scene. Right, fascinating. Right, and all of a sudden you have these genetic fingerprints a place where you've never been. My co-founder George Church, who's a professor genetics at Harvard Medical School, he insisted that we include,

if someone were getting sequenced, we couldn't ask them to get buy-in from all of their family members, but we could require that if any of them had a twin, an identical twin, that that twin would also have to sign up. Of course. That makes perfect sense. So, okay, was there anything else on what is possible right now on the personal genomics front? Oh, yeah. So in the present, you could argue that on the ancestry side,

we're getting much better at sort of getting a higher resolution view of who we are and where we come from and all of that. And it's an end of one example, but if you take the case of 23 and me, over the course of a decade has amassed a large enough genomic data set that it's clearly valuable from a research and development standpoint. It wasn't that long ago when they announced a, a collaboration with Glaxo-SmithK with GSC, where GSC is essentially paying them something on the order of $300 million

to get access to this data set to be able to drive some insights from it and potentially even follow up with people on a very opt-in basis. And so, you know, that will show you that at least on the original promise of person genomics, that's one example. That's been delivered, right? And so there is power in numbers. And I think the question like with any other technology, with any other resources, can we find the right balance where we're benefiting the commons

and not at the expense of the individual. And I think that's where a lot of the debate happens in terms of are we doing the right things. So that's where we are now. Let's talk about the future, since we're doing this whole Janus-themed episode. So given that there was this past of promise that was and wasn't delivered,

present of where we are, where are we going with personal genomics next? Or what is actually possible based on what we already know today? Yeah, well, I think there are certain things that are possible based on what we know today. The first one is, as these data sets become more rich, the ability to derive insights from them

that'll be relevant for how we diagnose or treat disease, I think that becomes increasingly more valuable over time. So what I mean by that is one of the big knocks on drug discovering development is that it takes a long time, it's very expensive, and the risk of failures high. One of the sort of lesser-known data points is that if you have a genetic insight driving a program saying that I think that a particular molecule or compound is going to be effective in a particular patient population that's defined by some,

sort of genetic or genomic marker, that molecule, that compound, that drug has a much higher, significantly higher chance of success. What does that mean practically? Does it mean that we can actually basically, is it natural extrapolation of that, that there may be a future where we do get personalized tailored medicine based on those molecules? That's right. The extrapolation of that is that we'll get better, faster, cheaper drugs that are tailored to the right population.

So they're like personalized cocktails at a mass manufactured level. Essentially, yeah, personalized cocktails. of therapies that at least are targeted to specific populations. That's actually the better way of saying that. Another potential future thing, especially we're doing a 10-year perspective look from today, is people talk about personal genomics. I think genomics is but one omic. Ah, yes, multi-omics.

That's right. Very big thing. If the big revolution over the course of the last 10 years is that we were able to sequence, you know, read DNA at a massive scale at a low cost, at high fidelity and all those things, that's increasingly true across many other ways in which, which biology transmits information. And I can bore you with all of the omics, but... Go through a couple of the hit list.

I mean, proteomics is one I know from when I was at park. So genomics is DNA. Proteomics is protein. Transcriptomics is RNA. Yep. Epigenomics is gene regulation and how genes levels are set. Metabolomics, the set of metabolites in your body. And of course, microbiomics and how the microbiome influences with all of that.

We're increasingly going to read biology across many, many frequencies. And by the way, we can also increasingly read biology at higher and higher resolution, which means you could read all of this information, not for a single individual, but increasingly from a single cell. And that's a very different thing, because now you could, for example, in a tumor, you could understand

how are the immune cells reacting to the tumor cells? How are the tumor cells reacting to the immune cells? If you can read biology at that resolution, we're going to learn a lot more about biology. Now, when you add to the fact that it's not just omics being transmitted by the cell that we can capture at high fidelity, but increasingly we have more censored data than ever before, more ability to crunch data than ever before,

ever before. I think if you look over the course the next 10 years, it won't be a question of personalized genomics. I think that will at some level be a dated term. It'll be the question of, you know, can you quantify individuals fully? And we're getting closer and closer to that. What would you say, though, I have to ask this because we don't want to be sitting here 10 years from now and asking, so what did we get wrong 10 years ago, Jorge, when you and I talked about this topics? Is it possible that Multihomics is also one of these much hyped things as well? I mean, we can't predict the future, obviously. Things play out. It's always a matter of timing sometimes when not if, where are we really on the spectrum of hype versus reality?

Yeah, it's a good question. I think if we take the last 10 years as any guide, there tends to clearly be sort of two stages, two phases, two ages. The first age is using technology to learn, and the second age is to use technology to act. If we look at what happened with personal genomics is that first age where we took to learn to gather data ended up being, I think, a lot longer than people probably originally anticipated. And we're seeing the benefits of how we can act on that, you know, towards the tail end of the last 10 years. I think it's probably reasonable to assume that if we look over the course of the next 10 years, to your question, the dividing line between hype and reality on something like multi-omics,

for me is really the dividing line and when do we shift from learning from information to acting on information. So one thing I wondered about, frankly, is the parallels. And your team talks about this a lot in terms of the parallels between engineering and the engineering phase coming to biology, which is that when it comes to detail, you know, DNA in genomics. The thing that's been most fascinating for me to watch for the last decade is that there is a Moore's law in genomics, and it's much faster than the regular Moore's law, and yet that pacing outcome of practical application is not necessarily on par with what happened with the

semiconductor. So that's where the analogy really breaks down, despite an accelerated effect. So one question for me is, what is missing in the ecosystem? Like, is it that there isn't the ability to manufacture? Is it a missing market, as you alluded to earlier? Is there missing components or materials. You know, when I think of the history of innovation, what still needs to be built out in addition to this core fundamental technology for this vision to come to reality? Ah, that's a great question. But first of all, I think the reason why we see a faster than Moore's law trend in genomics is because the ability to sequence and interpret DNA is really the confluence of three or four engineering marbles. You know, if you look at the next generation

sequencer, really what it is, you know, you're tracking the history and evolution of our ability to engineer better microfluidic systems. And other words, to move around tiny amounts of liquids. All about microfluidics from Xerox. There you go. So you're seeing improvements in the ability to engineer better chemistry. And this is both at the nucleotide level so we can get more efficient reactions. And at the surface chemistry of the platform, so you can actually run more and more reactions in tighter and tighter real estate. So you get more density. That's a second wave.

The third wave is we have massive improvements in optics. if you're going to run a bunch of chemical reactions in very, very, very small real estate, you need to be able to detect those. Optical detection. So when the reactions that drive sequencing are occurring at such density that they fall below the pixel detection level of the optics, you can't see the difference between them. So we had to see improvement in optics. And then all of that generated data that had to be deconvoluted with advanced computation.

And those are the four factors. And those are the four factors. Microfluidics, optical detection, improvements in chemistry, and data. Fantastic. So that's that revolution. So now why haven't we seen sort of the output look the same? The difference there is the output of Morris Law is better and smaller semiconductors. Those could be placed within a system that's been designed by people, like human beings, that could be optimized.

And therefore, you can get new products. Yeah. In the case of genomics, the output of that information has to go into a system that was not designed by human beings. It was designed by nature. So Jorge, bottom line it for me. So in this journey, from looking backward and looking forward, where are we in the personal genomics revolution and what should our takeaway be?

We're still in the early days of this revolution. If we look over the long course of time, we are still very much in the learning phase, in the data collection phase, in the information gathering phase. And it will be sometime before we make a mass shift into the taking action phase or into the productization phase

from all of this information. But when you look where we, are heading, that day will arrive. And that's why we are incredibly optimistic about what the future of genomics, of multilomics and biology more broadly will bring to our benefit. And when it comes to the privacy, in its full iteration, we will get the maximum power from genomic information when virtually everyone is sequenced. If we have perfect information, we can theoretically draw better insights. But that will come at important cost and considerations for how we treat the concerns of individuals that are contributing to that data

in a way that you're still protecting the individual but still benefiting the comments. Thank you for joining this episode. My pleasure. Thank you for having me.