a16z Podcast - a16z Podcast: When (and How) Biology Becomes Engineering

Episode Date: April 16, 2018

Hypothesis, test, revise -- that's science. Engineering, however, doesn't quite go that way: You have parts you know and understand (like legos), and then you use those parts to design and build something (like bridges). But the key is that when science -- time-consuming, unpredictable, slow, expensive -- becomes more like engineering -- faster, more methodical/repeatable, cheaper -- you can do new things... or do them in better ways. This means engineering disciplines like mechanical engineering, electrical engineering, computer science, and materials science can carry over to biology. But the question is HOW does this happen, and how can entrepreneurs apply principles from one discipline to another? How does it affect a healthcare startup's go to market, and how might a shift like this affect the healthcare industry as a whole? Vijay Pande and Jorge Conde (general partners on our bio fund) reflect on all this and more in this hallway-style conversation episode of the a16z Podcast, which was originally recorded as a video.

Welcome to the A16Z podcast. Today, we're having another of our hallway style conversations. These episodes are based on videos that are also available on our YouTube channel, YouTube.com slash C slash A16Z videos. What does it mean for biology to move from the high-risk, painstaking realm of the laboratory bench to the lower risk, get-it-done world of engineering? Bioteam general partners Vijay Pondi and Jorge Condé discussed this shift,

including how entrepreneurs can apply principles from one discipline to another and how it affects the healthcare startups go to market. Hi, I'm VJ Ponday, a general partner on the bio fund. Hi, I'm Jorge Conday, also a general partner on the bio fund. And so today we've got one of my favorite topics to go into. So, you know, we've seen companies that are built with science risk and companies, especially tech companies that are engineering companies, you know, how can we take things from the science curve,

which is stochastic and high risk, towards something that's more like engineering, more like grind it out, get it done, methodical, a great deal. less risk. Yeah, it's funny because we talk about this concept all the time. And, you know, reading a children's book to my kid, it was really interesting because science was defined as this concept. Yeah.

You have a hypothesis and then you test it and then you revise your hypothesis. That's some children's book, actually. Yeah. And then, you know, I picked up an engineering book at the Museum of Science and they defined engineering as this concept of you design something. Yep. Then you build it and then you test it and then you refine it and refine it and refine it. So is that what you have in mind when you talk about these differences?

Yeah, I think so. I mean, I think there's different ways to get there. So it's not going to be like a magic silver bullet. So let me just start with one and see what you think. So, like, one of the things I think we've been seeing a lot in biology is what I would call Legos. That you – biology is big and complicated, and you've got – but you got a natural hierarchy there. You've got atoms. You've got molecules.

You've got proteins. You've got membranes. You've got cells. You've got tissues. You've got organs. Got people. You've got organisms. You know, we can keep on going to –

the universes if we need to. You know, then the question is, can we find the parts? And if you can find the parts, then you can actually maybe mix a match. But do we have an example where that's already working in biology? Because the thing with Legos is you can line up the squares to the circles and make them click. Yep, yep, yep, yep. You know, that's actually a really important part of it. And so, you know, you can think of Cartier's like a simple example, right?

You know, that you've got two different pieces that you're putting together. You know, companies like Asimov is developing first the Legos and then putting those Lego, helping you put the Lego pieces together. And I think we're going to see more of that. It's hard, though, I mean, because it's not always obvious what the Legos are. And so that might be the science part of it. But then once you actually have the Legos, then you get to build stuff. Almost like, you know, when people build a bridge, they're not researching steel.

You know, they're given the girders and all the materials. And then they put a bridge together. So I think if people can come up with the parts, people can engineer from the parts. And do you think that so much of what has driven the industry, and specifically the sort of the biotech industry, we look at traditional space, is so much money and spend and therefore risk goes into trying to discover Legos? Yeah, I don't even know if they're thinking about it that way, right? I mean, because they're just trying to come up with a small molecule for the disease. And maybe what we're starting to see is like this shifting and this something, you know, you thought a lot about is, you know, what is a drug here? and as drugs become cells,

now the LEGOs become really important. Before, if drugs are molecules, your Legos are like phenol rings and things like that. And so there's not a lot to do there. But now there you have this big complexity. I think now maybe it starts become more important. Yeah, because I agree with you. I think one of the things that's interesting

is if we want to take this lens, and I'd love to get your take on this, if we want to take this lens of sort of science shifting to engineering, how do we think about high throughput biology within that context? Because one could make the argument that if you were going to do biology at massive scale,

at some point do you sort of stumble into engineering? Yeah. Yeah, I think that's an interesting point. And so one version of it is actually you could imagine two extremes. Or one to say, no, we're not going to do that at all because you don't do massive screening of bridges, right? You know, you design your bridge and then you're done. So on the other hand, I mean, in coming up with the parts, there's probably a lot of

experimentation. And so I could see roles for that. I think in this sort of parts analogy, I think the exciting thing is just once we've gotten there of what we can do with it, and that in many cases actually we've kind of gotten there. Yeah, I think that's right. And so if we look at engineering, you know, one of the neat things about this space as it's applied to bio, and I know you've spent a lot of time thinking about this and the firm and the fund increasingly and spending more and more time looking at opportunities here is what disciplines in engineering are applicable to biology.

Yeah, you know, that's actually something. really cool because I think it's a great point that, you know, when you typically think about mechanical engineering, you're thinking about literally bridges. But I think mechanical engineering, electrical engineering, material science, computer science, all these disciplines are pushing into biology. And so instead of steel, it's bone or its muscle. But the same principles actually carry over it really nicely. And if you look at on the academic side, academic departments, and they're really remadging themselves into these areas. And then the fruit of that's becoming an interesting fodder for these companies.

Actually, if we could pause on that point for a second. How is academia responding to this world, this sort of shift in the world? Because industry will follow, presumably. Yeah, I think, you know, one of the interesting things is that you don't have new departments created very frequently. Like, you know, physics departments are thousands of years old or something like, at least the Cambridge one is probably at least a thousand, something like that. So, you know, it's interesting that there is a new department that was created in the last 10ish years, the Department of Bioengineering, sometimes called Biological Engineering. And the creation of these departments, I think, has really accelerated this because you could have sort of inject people that are both engineers and biologists into engineering schools. And from there, I think, a nucleot's out.

And that biomegeneers sort of are having so much fun that I think other disciplines don't want to be left out. and that it kind of makes it easier to facilitate a straight mechanical engineering to get into this space. Yeah, you know, it's funny because I think one of the areas that I've been long interested in has been genetic engineering. Yeah. And if you look at genetic engineering as a field and as a practice, and I want to get your take on this, it was historically almost like it was genetic sciences, right?

Because you weren't actually designing. What was the engineering in genetic engineering? Well, it was kind of like plain boggle, right? You would just mix up all the letters and see what worked, when a word actually came out. Yeah, yeah. And now I think with the advent of things like CRISPR and things that companies like Asimov are doing what they're actually making DNA a design medium. Yeah. I think we're actually starting to make genetic engineering be an engineered discipline.

But I think that's a relatively recent. Yeah, there's an interesting point in that, which is basically just because you put the word engineering there, doesn't make an engineering. It's an aspiration more than a success. But actually, it's just even interesting that we have long had this aspiration. and that we're starting to see it. I mean, you know, we've talked about these disciplines. I think, you know, we talk about computer science. I think machine learning itself is becoming its own thing.

And it's maybe a final, sort of third way to sort of connect up with engineering. And that so much of science, when I think about it, is very bespoke that you come up with a biomarker. You've got some team of scientists spending years, maybe sometimes decades, to find this one marker. And to repeat that would be to repeat all that process and its stochasticity. Whereas something like machine learning is that you sort of engineer a process and then you just put new data and you roll through it. And so I think that might be sort of yet another avenue towards taking something that's normally in the science world and shifting it towards engineering world. That's actually a really interesting point because if I'm thinking about this from the perspective of industry and I'm a, you know, I'm at a biotech drug company and I look at my pipeline of drugs, you know, historically at least, the most advanced program is the most valuable one because that's the one that's closest to. hitting an inflection point, whether it's proof of concept or having data in a human clinical trial,

and the second one is the second most valuable one, and so on and so forth. So chronology determines value. And a big part of that, I think, was because of this concept of bespokeness, that you develop this first drug, and it doesn't necessarily educate you in what you're going to do with the second drug, unless you're going after the same target or the same biological pathway. But in your world, in the world you're describing this engineering world, it's actually the reverse is true, that the second drug is more valuable than the first, if you're using engineering, principles because what you learn from example one sort of imbues value to example two and so on and so and on top of that, you know, it's commonly asked why pharma doesn't look at their failures.

And I think the common answer is that, well, the feeling is that there would not be value in doing that. And that's why they don't put mine to it and be very expensive and so on. But if you're in this engineering curve, the false positives are actually as important to learning as the true positives. And so I think both of those get integrated in, which is just a new way of thinking about things. And it leads to a shift for, and I think we're seeing this more and more where farmer companies will start to view themselves more as data generating companies and data science companies as machine learning gets in. And machine learning often is like scary with AI and all this stuff. But if you think of it as just the best statistical use of the data, I mean, that's what farm has been wanting to do and trying to do for years. If you were to roll forward, maybe we'll pick a number 10 years from now, do you think a pharma company will have as big a dry lab, i.e. people on computers as they do a wet lab, I eat people at, I'd like,

bench with pipettes? Yeah, that's a really interesting thought. And in question, I think we're already seeing a little bit of that with just the shift to CROs where there's like not a purely medicinal chemist job, sort of drug designer job. And medicinal chemists have so much great intuition and experience design drugs that they've been sort of the tip of the spear of that. But as ML starts being competitive and hopefully really helping them forward, there may be a new sort of job, which is not doing the synthesis themselves.

but actually just be a drug designer, a drug engineer, throwing that aspirational term in there. And so that may be what's happening, especially if synthesis is done at a CRO somewhere else, then that design job, that engineering job actually becomes the real one. And it sort of shifts the tools you look for. In the end, though, the ML has to work, right? I mean, if it doesn't work, then that future doesn't exist

or is in a different timeline. So it's interesting because it occurs to me as you're talking that a lot of times when we talk about a company that has a platform, technology. Oftentimes that platform technology really is a combination of some technology, but really a lot of just know-how and expertise around a specific area of biology. Yeah. Yeah. But what you're describing in terms of this shift to engineering is platforms then can increasingly become companies, or companies that have platforms, I should say, increasingly become companies that have the ability to move drug, discovering development from being

a very bespoke thing. Yep. Yep. To being a very sort of productive thing. That's a good point. Like, ML is a platform amongst others. And then there's data-genering platforms or data analysis platforms. And that platform, I think, really could be a really interesting shift. And we're seeing more and more companies that way that are, you know, before it used to be that the platform was never really valued very much and it's all about the assets. And now we're getting to the point where people are very curious if platforms can reproducibly create

multiple assets. Let me ask you an unfair off-the-cuff question. You've seen a lot of things. what's the most surprising application of engineering in biology that you've come across, whether it's an academia or whether it's here on the investing side? Yeah. You know, it's hard to tell because there's so many different surprises. Like, you know, the favorite one I always love to talk about is like,

where we talked about this, that tree that glows at night, you know, the luciferous tree or something like that. And the idea that the reason why I find that so compelling is that it's a combination of sort of the technical engineering inside to make it happen, but also this idea that the future will not be steel and metal. It will be this sort of engineered biological thing that just grows, that has this function, something that actually helps fight against global warming, not contribute to it, something that really is sustainable, easily shippable because you ship them in little guys

and then they grow or something like that. It's just all of it is just a very different vision for what our world will be like. And it's something that it's just the beginning. And it maybe just starts with one tree, but then, you know, there's lots of different things. No, it's funny you say that because I think one of the things that's so fun about this concept of engineered biology is that, you know, at least in my sort of limited view, is, you know, historically when you think about engineering, it's making things better. Yeah. Making things more efficient. Yep.

Making things more quickly. But when you have biology as a design medium, it's about making things possible that you didn't even know where possible, like glowing trees. Well, we should probably talk about a little bit about nuts and bolts for how to get things done, you know? I think, you know, because it's nice to have the philosophy, but, like, how do you actually do it? And, you know, towards that, I think when I think about it, like my favorite paradigm is something like the Apollo mission. So President Kennedy says, we're going to the moon. Everyone, the engineer's like, ah, we're going to the moon. You know, I can only imagine that moment because that sounds crazy, right?

I mean, to do something like that. And how do you take something crazy, like going to the moon or, like, curing cancer or increasing human longevity by 50 percent? How do you do that? And the Apollo mission, actually the Apollo mission was even just the last one. You have Mercury, then Gemini, then Apollo. You know, as a kid, I really loved space. And just, you know, Apollo 11 got to the moon. So what was like 1 through 10?

You know, so one gets into orbit to, actually I forget all the details, but they have to practice docking in space. They have to practice all these things. They actually went around the moon before landing on the moon. And if you break it up into little bits, any little bit isn't so bad and can be engineered. And then you sort of do it step by step by step by step. And I think that's for me the inspiration for how to take some big crazy thing, like going to the moon, that if you did it from like a screening perspective to screen rockets to get over there, like that's going to be, you know, one a million chance.

But if you engineer little bits, bits by bits with, you know, OKRs, you know, your key milestones with KPI's, your key metrics, I can see how that starts to make a little more sense. But, you know, it's still easier said than done. but I think now it takes a big crazy thing into like lots of not so bad things. And that's important because that's where your Lego concept comes into play

because you couldn't build a rudimentary rocket if you didn't know where to put the screws to literally, quite literally pull it together. Yeah, that's a good point. Yeah, what they learned from, you know, the previous Mercury and Gemini

was to get the pieces and then they started mixing them together and even look at like the Soyuz rockets. They're like lots of rockets put together and so on. So, you know, in my mind that was one of those hallmarks sort of just landmark, like, wow, moments in human engineering. And I'm always curious to see how we can learn from that process to bring it over.

You know, with that said, there was a lot of people and a lot of effort in there. And so I think this is, again, not something that will be done by one company or one group, but, you know, I think collectively the ecosystem is starting to be built. So just to put this from the perspective of industry and from entrepreneurship, how does entrepreneur A get value for create value for creating Apollo 1? Yeah, yeah, actually that's a really interesting one, a tough one. I think there has to be this ecosystem from academia and research institutions to startups, to big startups, to big companies. And I think, you know, in Asmov's case, you know, a lot of the parts they came up with in academia for E. coli.

And then now they start to look at creating parts for other things. So they create the process. and now they can repeat the process for other types of cells. I think that might be part of it is that they got to like, at least to let's say the Apollo missions. Maybe they did Gemini, you know, at MIT. So something where you can get part of the way there where, you know, I think you and I can have some sense that, oh, we know where this is going and that this is now engineering, not like kids playing with rockets in the backyard,

you know, where maybe it's going to blow up or not and who knows it's going to work, something where we had the feeling like it's now we see the steps. It will also be interesting because I think, you know, there's lots of big incumbent companies, Google and IBM doing all this research, and they bring an engineering mentality and an engineering sort of zeal and excitement to other areas. And so I suspect we'll start to see that. I mean, okay, ours are like, I think part and parcel of how Google runs. And I would think that that would come over. Yeah, it's a good point because I think for companies that are trying to innovate in this space and using engineering as a way to do so in bio, the big challenge from a business. business development perspective has historically been proved to me if I'm the buyer, if I'm a large

company, prove to me that what you're doing is real and it's going to work. Well, yeah, so how does that work? I mean, how would you answer the question? I mean, what do they need to do? What would the buyer need to see? Well, so I think what's interesting, and we can take specific cases, Asimov is a great example where to use them. One of the beauties of what Asimov does is there's high predictability in what they design is going

to work. Yeah. And so if I, you know, if I'm on the, on the receiving end of looking at Asimov's technology as a potential partner, yeah, business development, a collaborator, if I say, you know, can you design something to do X? Mm-hmm. And it's significant percentage of the time it does, as intended. Yeah. That's very different than historical biology.

I think one of the big things in synthetic biology has really been, you try 10,000 things to get one thing to work. Yeah, yeah, yeah. And if somehow that ratio is, you know, nearly inverted, then all of a sudden it becomes a much more. it's much easier to get people to believe that you understand where the screws need to go. And if you understand where the screws need to go, then, as you pointed out, it starts to become clear that you'll eventually get to the moon. Yeah, so it sounds like you're saying that if you can demonstrate that you've sort of shifted onto this engineering curve, that would be sort of the key part for the go-to-market. Is that true?

I think that's exactly right. I think proving that something works as intended is historically been the big sort of activation energy you need to overcome for a buyer in the business development. And so when you're doing the stochastic risk that you're describing, that means there's a long period of time that has to go until you get this proof of concept. But what does that proof look like? Presumably it's more than just like a paper in science of nature or something like that. No, it's usually a, I'm going to show you X, you know, step-wide, stepwise improvement over a short period of time, or at least over a predicted period of time that what I can do works. Yeah. And that it's not a one-off, that it's reproducible, and that it works in various contexts. Because the various context usually represents some sort of a fundamental, if not universal truth.

So that's a proof of concept. That's like a small deal. I mean, how does that sort of actually? Yeah. So normally the way it plays out, it's usually a proof of concept that leads to a small deal. Yeah. And a lot of times, I think one of the challenges that early stage companies have is you don't want to run into this risk that you die from pilots. Yeah, yeah, yeah.

Right? Because I think the historically. Because they're easy, easy to get. It's easy to get into a pilot. Yeah, yeah. It's hard to go from a pilot to an actual agreement. And so this is where certainly the traditional biotech companies have had a very big challenge, right? Because every project is a science project and they're long and you don't know how they're going to turn out.

If you're doing something that's more on this engineering curve, you actually can, using your concept of OKRs and KPIs, you can actually say that this is going to be the project plan. And this is what we're going to deliver over a relatively short period of time because we can essentially iterate quickly. And if you can do that, then that leads to an initial project. And one of the things that certainly has been shown in tech time and time again is if you can demonstrate value at a small scale and prove that you can actually scale, then you have the famous concepts of the Land and Expand on the Enterprise side. You can actually start to see some of that in biology. The reason why Land and Expand historically hasn't existed in biotech is because the expand part was really hard because things were so bespoke. Yeah, you get your one drug and then you go to town on that one and then a platform was useful for getting you there. Exactly.

But then like you're there. Exactly. So how you focus on the drug asset? Yeah. So how does that change? I think it changes by this concept that, you know, asset number one is actually less valuable than asset number two, which is less valuable, asset number three, because you're learning from each one to the next one.

So you're getting better over time because you're moving from bespoke to design. And you're getting faster, you're getting like higher efficacy, lower talks, or all of those things? Presumably, yeah. So you're getting better, faster, and cheaper. I mean, that's obviously the sweet spot. Or in an ideal world, you're going from impossible. possible to possible. Yeah.

But that obviously is a higher hurdle. That usually doesn't happen in one step. That's right. That's exactly right. Yeah, yeah. But I do think that's interesting that when we see companies that have very powerful platforms, and there have been analogies for this in biotech, the ones that have very powerful platforms that can improve scalably and systematically on sort of an engineering-like curve,

yeah, tend to go from, you know, A, they take something that is impossible to possible over a longer period of time. Yes. And they tend to sort of grow and gain traction very predictably. because they can improve on a very predictable curve. And I know, you know, we use this example all the time, but it's what Aluminah did for sequencing. Yep, yeah.

And there are other examples where that has happened, where you've shown that you can demonstrate something in one context, you demonstrate it in a second context, and then by the time you get to the third, people just accept it as a general term. Yeah, well, that famous quote, I forget who it's attribute to the most powerful force in the universe is compounding interest.

That's right. It's probably Warren Buffett. And that's what this is, right? It's not money compounding. It's technology, like getting 30, percent better every year is, or 30-ish percent better, is like doubling every two years. That's like, we start with children's books. One of my favorite kids' books is that Indian story about the grains of rice, where like as a gift

or as a reward the peasant asked for two than four and doubling every day over a month. And the Raja thinks, you know, this is not that big of a deal, like two grains, four grains, it's not a big deal. And, of course, in the end, it gets to two to 32, it's like four million grains of rice, which was like all of it. And it just, it sneaks up on you. If you can get to that, that is how you make impossible possible. That's right.