Science is getting better at re-engineering micro-organisms for all kinds of uses, from better medical treatments to more durable materials. But there are still hurdles to overcome, including scaling the process.
Boston-based Ginkgo Bioworks was one of the first billion-dollar companies in the synthetic biology space. The NYSE-listed company uses machine learning and automation to coax biology to work at industrial scale. COO Reshma Shetty talks to Azeem Azhar about the company’s technology, business model, and how big she thinks synthetic biology could become.
@Azeem
@exponentialview
@reshmapshetty
Further resources:
The Bio Revolution: Innovations transforming economies, societies, and our lives – McKinsey, May 2020
Engineering Biology: The Next Frontier – Exponential View Podcast ft. Vijay Pande, June 2020
The Next Trillion-Dollar Market – Exponential View Podcast ft. Deep Nishar, April 2020
AZEEM AZHAR: Welcome to Exponential View with me, Azeem Azhar. The world is changing at an amazing pace. We’re entering the exponential age propelled by radical remarkable technologies. And on this podcast, I want to explore the themes, topics, and questions that will help you make sense of it. I want to turn chaos into clarity and help you make better decisions to make the right choices over the coming years, whether that’s in business, in policy or in your career. The exponential age has one certainty, and that is uncertainty. But over the next 45 minutes or so, my guest and I will try to shine just a little light on that darkness. And if we do that, please take a moment to share this podcast with a few friends. Part of what we are trying to do at Exponential View is make space for open discussion about how our world is changing, and that starts with you. This week, I’m talking to Reshma Shetty, the co-founder and chief operating officer of Ginkgo Bioworks. Ginkgo is using automation and machine learning to accelerate the process of designing microorganisms which could create valuable molecules that we can use in medicine, food industry, and indeed any number of other sectors. The potential market for synthetic biology is huge and Ginkgo is one of the leaders in the race to address it. Reshma Shetty, welcome to Exponential View.
RESHMA SHETTY: Thank you, I’m very happy to be here.
AZEEM AZHAR: And you’re definitely a biological organism.
RESHMA SHETTY: I can confirm that, yes. I’m even carbon based.
AZEEM AZHAR: You’re even carbon based, even better. The work that you are doing at Ginkgo was essentially a lab project for 10 years or so before you founded the firm. What triggered you and your co-founders to start a company?
RESHMA SHETTY: So, my co-founders and I met at MIT when four of us were graduate students, and then the fifth co-founder Tom Knight was my PhD advisor. When I showed up at a MIT, Tom painted this beautiful vision for me, what if we could program cells the way we program a computer? And I was actually a computer scientist by training and so this appeal from Tom really resonated with me. So, we spent our time at MIT trying to figure how would you program cells, how should we engineer biology? And candidly, it was hard. We weren’t very good at it, experiments took a long time. I could only make a handful of designs and usually they didn’t work. So, Ginkgo was in some ways born of that frustrations of a graduate student. We felt like there was another way to be doing this. We could leverage software and automation in a way that we just hadn’t been able to during our individual research and that was part of the motivation for starting Ginkgo.
AZEEM AZHAR: What’s fascinating is that lab was a computer science and AI lab rather than a biology lab. So do you think there was something about the fresh eyes of computer scientists looking at this domain that helped connect some of the dots?
RESHMA SHETTY: I think so. A lot of innovation ends up coming from the intersection of disciplines. If you kind of look at scientific history, some fundamental principles in biology were actually kind of discovered by physicists who came to the field, for example, like Jacques Monod. So I think Tom with his sort of engineering mindset, he was an electrical engineering computer scientist by training did have a fresh pair of eyes and had different goals, right? It was simply that his goal was not about discovering new biology, his goal was to engineer biology. And with that different perspective came different things he was optimizing on.
AZEEM AZHAR: We live in a world that is of course, full of biology and it is also full of engineering. But perhaps people haven’t thought hard enough about what it is to engineer something.
RESHMA SHETTY: So, fundamentally engineering is about design, right? I take something and I want to design it, I want to build it, I want to make something. So it’s a very tangible discipline, right? But biology had a very long history of being a discovery science. And it continues to stay and it’s amazing. There’s a ton to learn about the natural world and so that discovery science is really important, it’s the foundations upon which we rest. But fundamentally engineering is different. It’s about building stuff, about designing stuff, about programming stuff.
AZEEM AZHAR: There seems to be an interesting tension between the holistic nature of biology. One of the things that we have started to understand as we get into layers and layers below the sort of biological onion is that everything is connected. But engineering, we often think about as being quite reductive. Is there a tension there and how do you balance and match those two different systems?
RESHMA SHETTY: I think it’s absolutely true. Biology simultaneously operates at everything from the kind of atomic scale in terms of like synthesizing proteins and making new molecules to the level of entire ecosystems. So, I think that’s right, that if you’re then going to turn around and engineer it, you have to be able to kind of think and work across those different scales. We’re pretty good at engineering at the individual cell level. We can maybe do a little bit at the sort of organism level, like kind of, like maybe make tweaks but the fact is we don’t really have any idea how to think about engineering ecosystems. That’s a frontier that remains mostly unexplored.
AZEEM AZHAR: So is it the case that when we look at the opportunity of engineering biology, we actually are trying to start at the smallest level and we’re trying to figure out how to work our way up at larger scales in these more emergent dimensions that occur.
RESHMA SHETTY: Yeah. I think that’s fair to say. Like right now, a lot of our engineering efforts are focused on the individual cell, a bacterial cell, a yeast cell, a human cell but the hope is that as our tools get better, we at least have the ability to start to think about, should we be engineering at larger scales? But there’s a lot to discover and a lot to learn before we get there.
AZEEM AZHAR: There’s also a lot of opportunity out there. There was a recent McKinsey & Company report that said the opportunity for this new bio economy was many trillions of dollars. And I think when people think about biology, perhaps they’re thinking about medical product or food stuffs, but of course it’s much wider than therapeutics and vaccines. It’s also in many non-therapeutic areas like bio agriculture and animal proteins, in chemicals, in industrial enzymes and a whole range of other products that we get through a wide variety of different industrial sub sectors. So when we think today about what Ginkgo is building and working on, what are two or three of your favorite examples?
RESHMA SHETTY: So fundamentally Ginkgo is a platform company. So, our goal is to provide the enabling platform to companies across a pretty broad set of spaces to help them program cells for their application of interest. So for example, we work with a company, Motif FoodWorks. Motif is in the food space, and one of the things that Motif realized in part inspired by the success of Impossible Foods and their Impossible Burger, where Impossible, it had this amazing innovation where they developed essentially a burger that was made from vegetable proteins, right? But the crazy thing about the Impossible Burger is it doesn’t taste like your typical veggie burger, it actually tastes good.
AZEEM AZHAR: I’ve tasted one of them. I mean, it really does… is moreish.
RESHMA SHETTY: Yeah. I mean, you bite into it and it kind of tastes like meat. It even kind of looks like meat when you cook it, right? And the reason why it’s so different from most veggie burgers is that the Impossible Burger has in it a heme protein. And this heme protein, they got it from soy, I believe. It is actually what makes the Impossible Burger kind of taste like meat. This is really an impressive new product and it’s appealing to a lot of folks who are thinking about the environmental impact of the foods they eat. So, Motif sort of realized that this was the beginning of a trend where there’s going to be a lot of demand for these animal proteins to make new products like the impossible burger in the future. So, we’ve been working with Motif and they’ve recently announced their first product, Hemami which is a new ingredient for them.
AZEEM AZHAR: What difference has Ginkgo made to Motif being able to find that novel protein? Did you make it taste better? Did you help them find it faster? Did you help them find it more cheaply?
RESHMA SHETTY: We’re not experts in food science at Ginkgo. But Motif has built a team that is. They come from all across the industry and they know how to bring new ingredients to market. So, what we’ve done is normally Motif would have to build an R&D team, as well as that food science team to be able to bring an ingredient to market. But instead, they can leverage Ginkgo’s platform to able to screen a whole bunch of different ingredients, see which ones are both easy to make and taste good. So, we can kind of be that innovation engine for them and so that they can focus on what they’re good at, which is formulating and commercializing new ingredients. What we’re trying to do is essentially accelerate a lot of different companies across a lot of different spaces by offering that R&D engine to them.
AZEEM AZHAR: There’s another one that I quite like. I quite like this program with Joyn Bio which is all about the problem of nitrogen fixation. Now, I think the way I express this problem is that we need to have a working nitrogen cycle in order for us to have good agricultural crop yields. The way we have done that artificially for a hundred years is through the Haber-Bosch process. It takes loads of heat and hundreds of degrees Kelvin and hundreds of atmospheres of pressure and it’s basically a early 20th century controlled explosion and outcomes urea based fertilizers, which we then sort of spread liberally around the world and billions of people are alive because of it. And yet the natural nitrogen cycle which is done by little microbes sitting alongside plants takes place without there being hundreds of atmospheres of pressure and hundreds of degrees Kelvin of heat. Tackling that is a really important one from the perspective of climate change, not least, but also biodiversity and sort of soil health. And Joyn Bio is tackling this with you as well. What does that program look like?
RESHMA SHETTY: I think you nailed it in a nutshell, right? So it turns out as you mentioned, that there are a whole bunch of microbes out there that do fix nitrogen so they can take nitrogen from the air and basically convert it to a form that plants can use. So, usually what happens is that if there is fixed nitrogen around, say in the form of fertilizer, the microbes will just naturally shut down their own internal pathways for nitrogen fixation because they’re like, “Why should I bother? There’s plenty of fixed nitrogen around. I’m not going to waste the energy.” The challenge here is how do you get these microbes to keep fixing nitrogen through their biological fixation pathways even when there’s other sources of fixed nitrogen around? Because that’s the only way you’re going to be able to maintain crop yields, which is so critical for farmers to be able to deliver their crops at the cost point they need to. That’s kind of the core engineering challenge that Joyn is tackling.
AZEEM AZHAR: What’s fascinating to me is the breadth of these use cases. That was another example, which I’d love you to explain to me because I didn’t quite understand it, and this was a sort of a healthcare application thing called the Synlogic, which was going to help people who are missing certain types of enzymes. Help us understand that healthcare application.
RESHMA SHETTY: So Synlogic is a company based here in Cambridge, Massachusetts, and they really think about metabolic disease among other things. And in particular, there’s a whole set of folks who carry mutations in their genome, which means they might not be able to process a particular metabolite or they might make too much of a particular metabolite and it kind of throws off their entire metabolism, right? Oftentimes they have to be on pretty restricted diets because there may be certain amino acids that they just can’t consume too much of because of this sort of genetic mutation. The question that Synlogic asked is, “Hey, could we essentially supplement the metabolism of these patients with an engineered microbe?” Right? So the enzymatic functionality that they might be missing because of this mutation, could you essentially augment it or replace it with an engineered therapeutic microbe to treat the disorder?
RESHMA SHETTY: So for example, their lead clinical candidate is for a disease called PKU, Phenylketonuria. And the question is can you treat patients? And what’s interesting about this for me as a sort of a scientist and engineer is that this is a whole new therapeutic modality, right? There are just not that many treatments out there that are literally an engineered bacteria for treating metabolic disease. So Synlogic is really pioneering this whole new therapeutic modality and the hope is that these types of engineered bacterial therapeutics could be relevant for sort of metabolic and other gut diseases.
AZEEM AZHAR: What is the process of microorganism design and iteration? What does it actually look like?
RESHMA SHETTY: So cell programming is actually not that different from other engineering disciplines or even a cell software programming, right? Most engineering disciplines essentially go through a cycle of what we call design, build, test and learn and it turns out that that same cycle is the one that we have been applying to biology. So you have an idea for an organism or an application that you want to go after and you think about how you might achieve it, and oftentimes that involves a lot of research, right? Going and figuring out what’s known about the relevant biology, how has nature tackled this problem? Are there ways that we can take what nature has given us and improve it? So, there’s that sort of that early design phase where you figure out what might an organism look like that can make this protein or make this small molecule of interest? Then as you sort of sketch the high level design of the cell, you then go in and for each kind of step in the design, you’re essentially trying to design potential DNA sequences that would have that functionality, right? It’s sort of like if you’re a software programmer and you need to code up an algorithm for sorting a list, you go onto your computer, you open it up and you start typing it and spec out, here’s what I think a software program will look like to sort this list, right? We essentially do the same thing with biology, except instead of writing in a software development environment or writing in DNA.
AZEEM AZHAR: If I work back though, there’s a few different things that need to happen, right? So the first is the expression of what is the molecular output that we want, right? What enzyme do we want metabolized or what molecule do we want produced? So you have to be able to understand that and then you have to figure out what is the biological machine that can actually make that reference output? And then you have to figure out how do we program that biological machine? And the code that you are using is bits of DNA, which are presumably coming from the sort of gene libraries, right? We know that these genes are likely to express this kind of function, right? So there’s these sort of four different steps that work back from the reference design that we want and Ginkgo looks after all of those different pieces.
RESHMA SHETTY: That’s right. You can sort of think about it as like going from a high level like napkin sketch or conceptual idea and kind of drilling down the design hierarchy or the abstraction hierarchy all the way to what is the actual sequence of DNA that can encode this function of interest. So, we have folks here, cell programmers who essentially do that.
AZEEM AZHAR: And do they know the whole lot from beginning to end, or is it like in architecture when you’re building a house and the architect thinks about the big framing and ultimately the guy who puts in the lintels around the door is a real specialist, but isn’t going to be that helpful when you think about the electrics?
RESHMA SHETTY: I would say right now, most of our folks know about that whole chain, but we do have folks who sort of specialize with the higher levels of sort of architecting the overall project and how we might do it as you know architects for lack of a better term. And then we have folks who are really good at then taking that kind of spec and turning it into DNA design. That’s actually our design team. They’re genetic designers, they design DNA that they think will encode the function of interest. But the trick is that biology is not predictable, right?
AZEEM AZHAR: That’s what I wanted to ask, yeah. Right. So, what do you do with all this unpredictability and the kind of yields and statistical outcomes? How do you manage that?
RESHMA SHETTY: The way we manage it is by designing not one piece of DNA, but many pieces of DNA. So we’ll design thousand designs, 10,000 designs at a time and we’ll test them all out and see which one’s work and then learn from that process to inform the next round of design. This is not like even a computer chip where we have simulation tools that can tie in to tell us what works and what doesn’t. We have to go try it out in the lab. So, part of what we’re doing in biology and part of what Ginkgo’s platform can do is it affords the ability for people to be testing thousands of designs at a time.
AZEEM AZHAR: That seems to be one of the critical things, right? You talked about the design, build, test and learn cycle, which I guess we could try calling that DBTL and hopefully listeners will follow along. But if we look at the DBTL cycle in traditional software engineering where everything is quite deterministic, that cycle can actually happen relatively quickly and we have ways of scaling it out horizontally to do user tests, as well as sort of individual unit tests that does a code execute in a user test, does it do what people wanted it to do on mass? And I guess the challenge in the biological realm is that first of all, some of these processes just take time because the microorganisms have to work at biological speeds rather than silicon speeds and partly that you have these unpredictabilities right? The ways that ultimately genes express themselves is not necessarily a sort of one-to-one correspondence. Your conclusion then that the state that you got to is we have to build a system that allows us to eliminate delays, the end to end linear process, but then also parallelize what we do so that we can run enough experiments at the same time.
RESHMA SHETTY: Yeah, that’s right. Fundamentally biological design today is a search problem. For any given functionality that you’re interested in, there are many possible different sequences of DNA that may or may not encode that functionality. Because at every position you could have four possible nucleotides A, T, C or G, and you’re talking about genomes that are even IN the case of bacteria, a few mega bases. So it’s a very, very large search space.
AZEEM AZHAR: So how large is this search space?
RESHMA SHETTY: It is an unfathomable number of choices basically. We’re talking like as many stars as there are in a universe or type of things. Right now it’s just not possible to exhaustively explore that space. S,o what we have to do is try to bias our search to where we think designs are that will work. So, nature helps us a lot. Nature’s searched a lot of that space already. So, we can kind of go where nature has given us clues that there might be functional sequences and so that helps a lot.
AZEEM AZHAR: So it seems to be that there are perhaps three different ways of helping with that search, right? So one is that you use the cues and clues of nature and you build up a really big data set. And you say, if you want a protein that is tough and resistant and maybe a little bit shiny, it makes more sense to look at snails and Crustacea than it does to look at snakes or kittens and so you can kind of build a data set there. The other thing that you can do is you can parallelize lots of experiments. So a little bit like Darth Vader when he was searching for the rebel base in Empire Strikes Back, he sent out thousands of robotic space steroids, not just one, he parallelized his search. And I guess the other thing that you could potentially do is rather than things in vitro in actual wet lab stuff, you could perhaps start to do more synthetic simulation experiments perhaps by using some of these newer machine learning techniques that exist. So I guess first question is, are those the sort of three broad families of approaches you can use to make sense of this search space or are there others?
RESHMA SHETTY: Those are certainly three important families. I would say the other one that you didn’t mention but is also really useful is leveraging evolution. So biology is pretty good at searching this space for functional designs. So, you can also, for example, generate diversity and then use certain evolutionary methods to let biology find the working design for your application of interest.
AZEEM AZHAR: Oh, okay, that’s interesting. How does that work in the context of Ginkgo?
RESHMA SHETTY: So, for example, say I’m interested in a cell that can eat a particular nutrient of interest. And maybe this is not something that cells normally eat, but if you could get it to eat, like hey, it would degrade this waste product or it would provide a new carbon source, a new feed stock for your application of interest. Well, what if I essentially generate a bunch of diversity, there’s different ways to do that, but I’ll just say I’ll generate a bunch of diversity, a bunch of different cells with different designs in them, all of which are around a pathway for eating this feed stock of interest. And then what if I put those into a vessel and I give a limiting amount of that feed stock to the cell so that the cells that grow the best, the cells that can consume that feed stock, extract energy, extract carbon, they’re more likely to replicate and win, right? So, this isn’t always available to every application, but there’s sort of a lot of clever tricks that people have come up with over the years to sort of link up evolutionary pressure to a particular functionality of interest and so they can use selection to get to a working design.
AZEEM AZHAR: That is super, super interesting. I love these podcasts because despite all my research and the research workers that the team has done, I will still learn something. So you’ve got these four approaches then to help through the search space. We’ve got doing what ancient farmers did, which is the sort of evolutionary approach you’ve described, you’ve got sort of data sets and databases, you’ve got parallel experimentation and then we talked also about sort of simulation and sort of in silico approaches. When you kind of look at all of those, is it a case that you think that that whole portfolio is what’s going to be required or do you think that one of those approaches starts to become more and more important and deliver more results as we move forward? And hint of bias here, just wondering about whether it’s going to end up being simulation and being able to do these things increasingly in silico.
RESHMA SHETTY: I would say for now and for the near future, all four are going to be hugely important. Because the reality is that engineering biology is hard. There’s no one magic silver bullet technology that is going to solve it for us, even CRISPR CAST gene editing, which is one of the most amazing advancements we’ve had in the field over the last decade, two decades, even that is not the magic bullet to just solve cell programming. So, I think being able to bring all of these different approaches to bear on cell programming challenges is really important. I do think that as we get more and more data on what designs work and what don’t and we accumulate essentially what we call genetic code based, right? Which biological parts work, which don’t, which host strains are really good over producing this precursor of interest and whatnot, we’ll be able to shortcut some of that kind of high throughput experimentation that we have to do today. But I also think that as those tools get better, we’re going to expand the frontier of problems that we want to tackle. So yeah, there might be certain applications or problems that we can just do a handful of experiments and come up with an engineered cell to produce, but then we’ll go tackle the next hardest thing, which will then once again, require all four of those approaches to be able to realize that application of interest. So, I see it as sort of an expanding envelope of what is possible as the technologies improve.
AZEEM AZHAR: I’m curious about this idea of genetic code bases or biological libraries because there have been a number that existed before Ginkgo was founded. There’s the genetically engineered machine, the GEM libraries amongst other things that those earlier efforts didn’t seem to result in sort of huge industrial shifts. What do you think is different about the kind of approaches we’re seeing today than the ones that perhaps were taking place 15, 20 years ago?
RESHMA SHETTY: I think one of the challenges in the industry is that many companies were for very good reasons, product focused, right? So they said, “Oh, I want to make this more molecule of interest and I’m going to go accumulate these genetic assets to be able to go make that product of interest.” But fundamentally they were only being used for that one product of interest. But that was like kind of the most valuable property and assets of that company and so there wasn’t much incentive to share those with any other company. So, if you had a totally orthogonal application that might benefit from those same genetic assets, it’s actually pretty hard to get your hands on it because that’s the property of this other company.
AZEEM AZHAR: So, this kind of transverse horizontal learning across projects wasn’t possible in a way.
RESHMA SHETTY: Exactly. There’s essentially a Balkanization of these parts libraries across a bunch of different product focus companies. So, the bet here is that at Ginkgo, we can start over time, accumulate more and more of these assets and enable their reuse across very different applications, very different products and across all of our customers.
AZEEM AZHAR: Through this process of cell programming, we have now got our alpha cell, the thing that does what we wanted it to do. It’s only one cell, it’s not going to change an industry. So how do you scale that up in manufacturing terms? What is the way in which we make many of these things so they can actually be useful?
RESHMA SHETTY: So the amazing thing about biology is that it can self replicate. So, essentially cells are capable of making copies of themselves and growing to actually very large scale. So, the way we scale up production of a lot of different products from our engineered cells is that we use a process that’s basically not too different from brewing beer. So you grow up yeast in a Vat and as they grow, they’re producing a product of interest, in that case, they’re making ethanol for beer, but in our case, they might be making a protein of interest, a chemical of interest, or you might be growing it up to use the microbes themselves as the product in the case of Synlogic and their engineered therapeutic bacteria. So, you just kind of grow them and Vats and feed some sugar in or other feed stock, and then produce your product of interest, and so then typically there’s some sort of harvesting and purification process that you use to actually purify your end product from that fermentation.
AZEEM AZHAR: This bio manufacturing requires these fermentation tanks. One of the things that struck me is that, of course we have fermentation tanks. We have large ones that are used in the food and drinks industry and we have these really small ones that are used in the pharmaceutical industry. The ones for the pharmaceutical industry are higher grade, right? They’re easier to control, high purity and so on. One of the things that struck me about your business is that a lot of the outputs that you need are required at very large scale, but with the sort of precision and purity of the typical farmer style outputs. What does that mean for you? Did you have to design your own fermentation tanks, fermentation systems, or are there suppliers, or is this just something that will evolve in the industry over time?
RESHMA SHETTY: So there are manufacturers out there called CMOs, contract manufacturing organizations that actually have spare fermentation capacity. Sometimes you can go to third parties who do this, also sometimes our customers have this capacity themselves in house. So, essentially what we do is we usually have some idea of where we’re going to be making the product of interest when we’re designing the organism. So, we try to design both the organism and the process, how we culture it, how we feed it, how much oxygen we put in and whatnot based on where it’s going to be manufactured ultimately commercially. And we try to design that process to be as sort of commercial scale manufacturing friendly as possible. So, we put that upfront work in to help optimize the likelihood or increase the likelihood that things will go well at commercial scale manufacturing.
AZEEM AZHAR: I wonder whether this sort of commercial scale manufacturing is as easy as perhaps sometimes it’s made out to be. Because of course, biological cells do replicate themselves, so that’s why we’re all here. But we also know that that replication can go wrong. I mean, that is the way of the cancer and the tumor. What part of the DBTL testing that comes out with a single microorganism naturally guarantees that when you start to get these things to replicate by the trillions, the output you get is the one that you want consistently. So how easy has it proven to be to get to this sort of manufacturable scale?
RESHMA SHETTY: Getting to commercial scale manufacturing is actually quite challenging, but there are certain things you can do to again, increase your odds of success. So, for example, I can actually control how many replications happen in the course of my manufacturing process. I can build what’s called… something called a seed train, right? Where I have seed stocks that I’ve kind of Q-seed and validated and I know that as long as I maintain only a certain number of replications as it goes to commercial scale, it should work. The DNA should not break on me, right? That’s like one example. Another thing I can do is test how robust my cell and my process is. So when I’m at small scale, I can look at, hey, what happens if the pH suddenly goes wonky? Or what happens if there’s a different level of oxygen than I was expecting? How robust is my cell and my process to those, what are called process excursions? So, part of my DBTL cycles that I’m running are not just to get to a working organism, it’s getting to a robust working organism and process. As I’m something that we think we’ll be pretty tolerant of that type of commercial scale manufacturing variability that we might see, and so sometimes the best microbe isn’t the one that makes the most. It’s the one that makes a pretty good amount, very consistently despite any variation there might be at commercial scale.
AZEEM AZHAR: It’s one of the things that I was always wondering about, right? When you start to brew these things in 50 or a hundred thousand liter Vats, these microorganisms need particular conditions, but the temperature, the level of oxygenation, the pH, the fluid dynamics across that large volume is not going to be precise, it’s going to vary. There are going to be hotspots. There’s going to be a process of equilibration going on. And that struck me that that’s quite hard to control. And can these little critters actually manage it and do the thing we’re asking them to do?
RESHMA SHETTY: That’s sort of part of what you’re screening for when you’re doing your design, build, test, learn loop, is you’re looking for microbes that will keep producing even with that variation in environmental conditions.
AZEEM AZHAR: What are you learning about how fast each of these stages is compared to the approaches that we were using before? How much more quickly are you able to design a microorganism compared to the pre Ginkgo world?
RESHMA SHETTY: The design, build, test, learn loop is still a little bit longer than we’d like to be honest. And that’s part of the reason why we parallelize the designs, is because the length of that loop is longer than we’d like and it’s also can be hard to compress, right? Certain steps, for example, building up short pieces of DNA into longer pieces of DNA, it just takes a while. Still kind of frustrating, but it takes some time, and so that can take a few weeks, right? So that’s pretty long, right? That’s a long step in your process and it’s been hard for the industry to kind of compress that. Second thing is cells need to grow, right? They have to take a certain amount of time to grow up, and so that kind of creates, again, another fundamental limit on the length of that cycle. So, while we can kind of get rid of lag times and tighten it up and there’s more to go there, there are going to be certain fundamental limits beyond which you can’t compress that design, build, test, learn cycle anymore. So, that’s why both parallelizing it as well as leveraging some of these other methods like learning from nature or evolution or computational design is so important.
AZEEM AZHAR: I mean, you draw a lot of parallels with the traditional semiconductor industry in the idea of fabs and foundries and this iterative design process. And one of the key things that has driven that industry forward has been its ability to miniaturize as well. I mean, to what extent is miniaturization a strategy that you’ve been able to take advantage of to improve throughput or to use timescales?
RESHMA SHETTY: Miniaturization has been important for us, but the reason it’s important is one for cost reasons. So, some of these reagents we use when we’re either building a DNA or testing cells, they’re expensive. So, if you can miniaturize your assay, you use less reagent, you can reduce the cost of any individual strain design to extend it through the cycle, which means you can, again, test more designs in parallel.
RESHMA SHETTY: So cost is a big driver for us on miniaturization. Sometimes it’s also actually faster, right? So moving large amounts of liquid around turns out to be slower than moving small amounts of liquid around. So, miniaturization can actually give you some benefit on the speed side as well.
AZEEM AZHAR: You’ve introduced several different disciplines already from genetics to brewing, to machine learning, data science, robotics, fluid dynamics. How different are the philosophies of the software engineers working in Ginkgo to the biologists, and what do you do to get them to work together?
RESHMA SHETTY: Everyone who’s here is here because at some level they believe in the mission. So at Ginkgo, our mission is to make biology easy to engineer, and that’s sort of what ties everyone together. But you’re right that we have a whole bunch of different disciplines here and including pure play biologists all the way to kind of hardware engineers or software engineers. I think the biggest kind of barrier to overcome there is can oftentimes be one of vocabulary. Because the words we use when we talk about programming software are different from the words we use when we talk about engineering biology. So, that kind of vocabulary and sort of background understanding is an important bridge we have to try to cross within Ginkgo.
AZEEM AZHAR: So, that’s the details of the internal process that I think we can capture for now, but of course you’re working with customers. One of the things that’s fascinating about Ginkgo is that you have a platform business model, you’re not out there producing a single end user product, and that means that you are working with customers as partners to develop these new molecules. Now, in many cases, the way you work with them is in a joint venture sort of structure style approach. What are the determinants of which customers you work with and how you actually structure those relationships?
RESHMA SHETTY: Because we’re a platform company and because we see certain economies of scale in our platform, like the more cell programs that are happening on top of our platform, the more all of our customers benefit. One because of the scale economic, right? The more work that we’re putting through our platform, the lower our unit costs, but also because there are learnings, right? Every cell program we do, we learn something. We develop new biological parts, we develop new data, new genetic code based assets that can be reused across cell programs, and so we think that all of our customers benefit the more cell programs we have on top of our platform. So, because of that, we have been pretty flexible/creative in the types of deal structures that we do because different customers have different needs candidly. If we are working with big multinational company, then they might be pretty interested in just using cash when they work with us, right? Because they might have a lot of it and that’s the easiest way for them to do a deal with us. But for other companies say if you’re two grad students just starting a brand new company and you want to kind of build a prototype to see if you can get off the ground, you might be really cash strapped, right? And then sometimes we’ll work with them and for example, with structures where we might get some mixture of fees and equity. So we tend to be pretty flexible.
AZEEM AZHAR: What is the qualification mechanism that you use when you are talking to a potential customer?
RESHMA SHETTY: The first is simply technical feasibility. Like, is this a project that we think can be done? Two is the financial one, does the economics of this deal make sense and for us as a company? So typical like any company would do. But then a third lens that we bring to the projects that we do is one that we call caring, right? So we’re a platform company. There are a lot of different cell programs that we could do, a lot of different customers we could work with, right? But we haven’t explicitly taken a stance that we are not values neutral when it comes to what cell programs we will do on top of our platform. We’re clear how our platform is used. This is a bit in contrast to how some other technology platforms have developed, where they tended to take more of a position of being neutral with respect to what content or what work or what people are on top of their platform. We have decided and taken a position at Ginkgo that given the power of biology, given the power of this platform, we are not going to be values neutral when it comes to what cell programs we’ll take on. So for a lot of programs, this may not even be a factor, but we do tend to apply this caring lens to what cell programs we do on our platform.
AZEEM AZHAR: So if I wanted to come to you to bio-engineer some sort of really aggressive virus, you might say this doesn’t meet our caring threshold and we’re not willing to work with you.
RESHMA SHETTY: Correct, yeah.
AZEEM AZHAR:You’re obviously well aware of the widespread debate that’s been going on in technology circles, software, internet companies around this issue of values. And it’s great to hear you say that, that as a firm, you ask those questions of yourself. But of course it begs some additional questions, which is how do you go about doing that and what were the precedents that you learned from in terms of this sort of decision making?
RESHMA SHETTY: Candidly, I think we’re as a company still working through, what does it mean to care how our platform is used? So, we learn from every particular case and try to get better at this over time. But I think there’s a few rules of thumb. One is that we actually think that diversity ends up being really important to care, right? With platforms, there can be a lot of unforeseen consequences or unintended consequences from how that platform goes out into the world. So, we think that if you have a more diverse group of people who are building the platform, who are using the platform, who are selling the platform, then you’re more likely to be able to see around those corners. So, a lot of our internal diversity, equity, inclusion efforts essentially have a business motivation behind them candidly, because we think a more diverse workforce will be in a better position to hopefully see around those corners.
AZEEM AZHAR: Does that mean that you have this sort of fierce contest, there will be an argument around whether a project should go in a particular direction or whether a new client should be taken on because of perhaps their heritage or the sector in which they’re trying to operate?
RESHMA SHETTY: Yeah. There are cell programs where there’s not a clear cut answer about whether or not you should do them and so we have those internal discussions, yes.
AZEEM AZHAR: On the customer side, it seems like you are often taking quite a lot of risk in terms of the customers that you choose to partner with. There’s a bit of a venture capital aspect of this because there’s a lot of risk about whether you’ll ever be able to find that organism over a given timeframe and whether you’ll be able to manufacture it and then how successful the new ingredient will be wherever it’s applied. So is there a bit of a sort of venture capital aspect in terms of who you partner with?
RESHMA SHETTY: I think it’s certainly true that there are sort of risks of different steps in the process, but the way we manage those risks is like venture capitalists with a portfolio approach, right? We work with a lot of different companies across a lot of different spaces and a lot of different applications, and so we’re not naive enough to believe that every single one of our cell programs will ultimately be commercially successful, but the bet we’re making is that enough of them will to justify the entire enterprise.
AZEEM AZHAR: Is it the case that when you work with customers that Ginkgo holds the intellectual property of these particular designs and if so, what drives that choice? I mean, it seems to be something that favors Ginkgo more than the customer in that respect.
RESHMA SHETTY: So one of the issues that we observed in the industry is the one I’ve alluded to, which is that most companies in biotechnology have been product focused companies. So, they develop a set of potentially really useful genetic assets, but then they end up only pursuing one product or a handful of products. To me, that’s a lot of wasted potential, right? Because those assets are probably useful for other applications. They’re not even necessarily competitive with the products that a particular company is interested in, but oftentimes those assets lie fallow or don’t get use, don’t get fully developed for all the possibilities. So, one of the things that’s been important in our model is that we want to avoid that, right? Avoid that wasted potential. We try to work with customers to negotiate IP terms that benefit both where they can use it for their product of interest, but we can also reuse it for other products so that again, all of our customers, everybody who’s working on top of our platform benefits from the cumulative learnings and intellectual property that have ever been developed on our platform.
AZEEM AZHAR: And I guess if a customer’s not comfortable with that agreement, there are other choices they have, right? So that’s in terms of commercial.
RESHMA SHETTY: Correct, yeah.
AZEEM AZHAR: Right? Fair enough. Now your co-founder Jason Kelly has sort of said that we’re at the end of the beginning for synthetic biology. So what’s next?
RESHMA SHETTY: I think the going’s just getting good when it comes to the ability to engineer biology. And we’ve been investing a lot in the platform over more than the last decade, right? And we’re starting to see more and more cell programs come to technical fruition. My hope is that we’re going to sort of see a snowball effect where as people see these proof points out in the marketplace for products that have been made via biology, it’s going to just drive more and more interest among folks in leveraging biology as way to solve their problems. And one of the ways we’re seeing that is candidly on the startup side. We’re seeing more and more startups get started who have a application in mind and want to use biology to solve their problem, and so they’re excited to kind of sit on top of Ginkgo’s platform to be able to pursue their idea.
AZEEM AZHAR: And what are the main hurdles then for some synthetic biology to fulfill its promise? Is it that on the customer side, not enough customers understand what the potentials are, is it that the discovery process is too complex or the scaling and manufacturing process needs to improve?
RESHMA SHETTY: Well, I think the answer varies depending on how complex of a engineered cell that you’re trying to make. So for example, if you’re interested in engineering a cell to produce a particular protein of interest, we’ve gotten pretty good at that. There’s some technical risk, but not a lot, particularly compared to where we were 10, 15, 20 years ago. Our risk for profile has been shifted actually more to, can you find the protein of interest that has the properties that you want for your particular application and can you go formulate it, market it, sell it in a way that folks will find compelling? In certain areas of cell programming, the risk has shifted out of the kind of technical realm and into more of the commercial and marketing risk. In other areas we’re still trotting new ground when it comes to the science and the biology, because it’s a lot more complex and so there remains more technical risk. But as the technology platform gets better, as we accumulate more of those biological learnings, I think those two hopefully will shift. So, again, I sort of view it as an envelope of the frontier of which is growing it with each year.
AZEEM AZHAR: Well, one of the challenges with a magical technology like this is the hype cycle, right? There’s a risk of inflated expectations, and in particular that can sometimes mean that people in the market don’t understand this. I mean, how well do the capital markets understand Ginkgo?
RESHMA SHETTY: That’s a difficult one for me to speculate on. I would say that we went public in September of 2021. And through that process of going public, we’ve been out telling our story and telling our story in a probably much more like full some and detailed way as we became a public company. So, I’ve actually been pretty surprised by how well that story has resonated with the markets, but there are certainly split opinions out there about whether the time has come or not. So reasonable people can disagree.
AZEEM AZHAR: We’re going to run short of times so I just have one last question for you if I may. What would be the moonshot dream application, the one that you would say, tell your parents and your relatives and all of those around you that you had the greatest pride in helping to develop? What is the dream application that could come out of a platform like Ginkgo?
RESHMA SHETTY: So, I dream pretty big. My moonshot is not a moonshot, it’s a Martian. I would say terraforming Mars might be the ultimate application of biology.
AZEEM AZHAR: Well, Reshma Shetty, thank you so much for your time. Maybe the next time we speak, you will be Zooming in from Mars in a nicely terraformed planet courtesy of one of your or several of your microbes.
RESHMA SHETTY: Oh, one can only hope.
AZEEM AZHAR: Thanks very much. If you enjoyed this podcast, please do consider passing it on to a friend. Part of what we’re trying to do at Exponential View is to make space for open discussion about how our world is changing, and that starts with you. To become a premium subscriber of my weekly newsletter, go to www.exponentialview.co/listener where you’ll get a 20% off discount. And to stay in touch, you can follow me on Twitter. I’m @Azeem, A-Z-E-E-M or A-Z-E-E-M. This podcast was produced by Mischa Frankl-Duval, Fred Casella and Marija Gavrilov. Bojan Sabioncello is our sound editor.