NEB Podcast Episode #8 –
Interview with Tom Knight: The Future of Synthetic Biology
Interviewer: Lydia Morrison, Marketing Communications Writer & Podcast Host, New England Biolabs, Inc.
Interviewee: Tom Knight, Co-Founder of Ginkgo Bioworks
Hey there, welcome to the Lessons From Lab and Life podcast. I'm your host, Lydia Morrison, and I hope that our podcast offers you some new perspective. Our podcast today focuses on the field of synthetic biology. I'm joined by Tom Knight, who started taking courses at MIT when he was just 14 years old, and later helped design the first bitmap displays and ITS operating system. Then, he decided to apply what he knew about electrical engineering and computer science to biology. He's one of the co-founders of Ginkgo Bioworks, a synthetic biology company in Cambridge, Massachusetts, developing new organisms that replace technology with biology.
Today I'm joined by Tom Knight. Thanks so much for joining us today, Tom.
Thank you for having me, it's a real pleasure.
Meet Tom Knight, The Godfather Of Synthetic Biology
I read that you have been called the "godfather of synthetic biology", can you tell me how that nickname came to be?
Well, by background I'm a computer scientist and an electrical engineer, and spent many, many years at MIT. Got all of my degrees there, actually came to MIT as a high school student. I was brought up in Wakefield, Massachusetts, in about 1990 or so I realized that the next great technology was not going to be based on silicon or electronics, but was instead going to be based on biology, and that if I was going to be technologically relevant looking forward, that I had better be part of that. So, kind of took a right hand turn technically at that point, and basically became a biology student at MIT. So I took the undergraduate laboratory courses, I sat in on many of the graduate courses in biology, focusing almost entirely on prokaryotic biology.
Approaching DARPA about Cellular Computing Funding
Then in about 1996, I approached our funding agencies at DARPA, and DARPA at the time was heavily invested, and still is heavily invested in computer science research. Most of our computer science research had been done, sponsored by them. I approached those computer science oriented people and made the pitch that they should be thinking seriously about biology looking forward, and they run a summer program at the national academies in Woods Hole.
And that summer program looks at advanced technologies and tries to evaluate which of those technologies the DARPA offices should be funding, and I did a study with them in 1996 looking at, what at the time we called cellular computing, and the focus was should we be thinking about basically an engineering approach to biology? And that's where I think this field really took off, because it was really the first time that people had thought about building a technology on top of biology, something that could do things that you might expect a computer or other kinds of heavily engineered and technologically oriented devices, how you could make those out of biology. So that really is where things happened, and I think when people look back and they, many of them, I think myself included, view that as really being the place where the technology started.
Early Years in Computer Engineering and Semiconductor Design
That's really interesting. So that was over 20 years ago. It's really amazing that you were able to make that leap from computer programming and semiconductor design and think to apply that sort of logic to biological systems. So, thank you from the scientists, I think, for making that leap to transitioning to a new field. Do you think that you would have been able to come up with an idea like that had you not spent decades learning the ins and outs of computer programming?
Probably not. I think historically I have always been somewhat of a scientific and engineering generalist. I've always had a variety of interests. In high school I thought I was going to become an organic chemist. So as I went through the different areas of computing, many of those involved fairly serious transitions, from programming in Fortran, to starting to program in machine language and writing operating systems, and then finally into the areas of computer architecture and designing computers, and then really a pretty radical change that started in around 1980, moving into the design of integrated circuits and the silicon technology. All of those involved fairly radical shifts in the kinds of things that I needed to know, and it was very important to have an ability to move quickly from one discipline to another. So this was just the latest one.
How long did it take you, do you think, do get up to speed in biology? You mentioned you sat in on some undergraduate courses, and some graduate courses. So how long does it take to learn to become the master of a new field?
My gosh, I think I'm still learning. I don't make any claim to be an expert, although I hold my own sometimes.
I'm sure you can.
But I would say in about 1997, as a result of that DARPA summer program, they funded a molecular biology laboratory in the computer science department at MIT.
So I had the experience of setting up a lab. That was a learning experience, as they say.
So I learned a lot about what it took to actually do these things. Some of my colleagues in the computer science department I'm sure thought I was going to kill them, but they seem to have all survived.
Founding of Ginkgo Bioworks
Well, thank goodness for that. So about 10 years after that DARPA meeting, you founded Ginkgo Bioworks, or co-founded Ginkgo Bioworks.
Can you tell us about what Ginkgo Bioworks does?
From the very beginning I had the privilege of working with four of my co-founders, Reshma Shetty, Austin Che, Barry Canton, and Jason Kelly. Two of them were my students, PhD students, and two of them were Drew Endy's students, and Drew is now at Stanford. We worked very closely together in the early 2000s, and when they were all graduating, roughly at the same time, completely at the same time in 2008, Jason approached me and said, "Well, what would you think if we just started a company?" Rather than going to try to find an academic position or a lab somewhere to do a postdoc. And I said, "That sounds great, can I join you?" Because frankly, at that point I was pretty fed up with the academic world. It had become increasingly difficult to get funding for the proposals that I wrote.
I wrote what I thought were extremely good proposals, and the funding people in many cases told me that they really liked these proposals, but at the end of the day there was not money coming in the door.
Do you think they were a bit too forward thinking and far reaching for the types of funding that were available?
Possibly. That's been a problem of mine for many years.
Thank goodness for that.
Ginkgo Bioworks Mission
But in any case, the goal of Ginkgo was, and really still to this day is, best exemplified by the mission statement of the company, which is to make biology easier to engineer. That fundamentally is what guides us, it's what we think about on a day to day basis, and how we go about doing that and the methods and approaches that we take, those all change, but that fundamental direction has never really changed. So we're focused laser-like on can we move biology from being a scientific discipline to being a high-throughput technology, something that you can build upon, a foundry based approach to biology. Much in the same way that the semiconductor world has developed semiconductor foundries where you have a variety of different companies, some of them who have no access whatsoever to semiconductor fabrication, are doing designs of semiconductor components, but then they're fabricated in a central facility.
We had much that vision for Ginkgo. We wanted to build a technology which allows a variety of different projects, or a variety of different customers to come together to share a centralized facility that is able to do extremely effective engineering of biological systems, and that's what we've tried to create.
New Projects at Ginkgo Bioworks
That's really interesting. I'm sure there are propriety restrictions, but can you tell us about any of the interesting ongoing projects that you're working on?
Well, there's a couple that I think might be easy to talk about. One particularly interesting one that attracts my attention, we made early efforts to go into the flavors and fragrance area. So we have developed close relationships with some of the perfume manufacturers and are doing a number of projects for them. I really can't talk about those projects, but we are doing a fascinating project in collaboration with the Harvard herbarium. The Harvard herbarium over the past 200 years or so has collected samples of flowers from all around the world. Some of those flowers that they have collected are now extinct.
So we have a collaboration with the Harvard herbarium where we can get access to dried samples of flowers, we can sequence the DNA from those extinct flowers, and then using our high-throughput DNA synthesis we can make the genes that are pulled out from those extinct flowers, and recreate some of the enzymes that might be found in those flowers. You can't, of course, know that the compounds that you're making are ones that were really made by those flowers, but you can make a pretty good guess that the compounds, the scents that are being produced by those terpene synthase molecules are in fact representative, at least in a crude way, of some of the scents that those flowers might have produced.
Collaboration with Bayer Crop Sciences
That's so interesting to think that you could smell the scent of an extinct flower. What an interesting way to preserve and to add to that catalog of flowers.
Another project that has been widely announced is our recent collaboration with Bayer Crop Sciences, where Bayer, the large chemical manufacturer, has for many years had a presence in the agricultural business. We were chosen by Bayer to set up a joint venture between Ginkgo and Bayer to make it possible to develop a nitrogen fixing bacterium that they could apply to seeds in the field, and the idea would be that you could take crops like corn and wheat, that naturally do not produce their own nitrogen, to the extent that soybeans and peanuts do for example, you could take bacteria that are able to fix nitrogen from the air and replace some of the chemical fertilizer in the field for these very common crops with microbes, rather than chemical fertilizers.
Interesting. So that would aid in fertilization and growth?
Would remove the need to do at least as much of the chemical fertilization, which has a whole set of environmentally dangerous aspects of run off of nitrogen fertilizers into water supplies, and so forth. Also, of course, has a cost impact.
So these would be natural production through microbes of nitrogen?
Yes, but engineered microbes.
Right, not naturally occurring microbes.
Not natural microbes.
That's so interesting. When will we expect to see those crops in the field?
Well, it's what we sometimes call a moonshot project. It's recognized I think by everyone to be rather difficult, it's a five year program, and so I think there's no reason to expect that it's going to be next year.
That's so exciting though, and really nice to hear that efforts in biological engineering are geared towards reducing environmental impact of agriculture.
Advances in Metabolic Pathways
Could you tell me how recent advances in technology, and the introduction of robotics has moved forward your workflow or your pipeline?
A typical project that we engage in involves assembling pathways of several enzymes to produce final products. The approach that we're taking is, as a result of the high-throughput screening that we have, uses some rather different ways of approaching that process.
A typical pathway that we design looks at each of the enzymes in that pathway independently and typically when we start the project we know at least one enzyme that is capable of catalyzing a specific step in that pathway. The approach that we take is to use the amino acid sequence for that protein and search our databases, both the publicly available databases, and also our proprietary internal databases, and pull out hundreds of thousands of amino acid sequences that are similar to the enzyme that we start with. We may also pull out enzymes that are predicted to do similar reactions, but which are not particularly related at a sequence level with the enzymes that we start with. We then construct a phylogenetic tree of those enzymes, and it's typical to see that a large fraction of them are very, very similar. They may come all from sequencing E. coli, for example.
We choose representative samples sparsely from that list, and pull out perhaps a thousand versions of amino acid sequences that we think are worth testing. Those amino acid sequences are then recoded into DNA sequences for the specific organisms that we're planning on using, and are handed over to our synthesis team. That synthesis team will then take those sequences, use high-throughput techniques to synthesize all of those fragments, and then our transformation team will insert those sequences into the target organism, will grow the organisms up, and will test the effectiveness of those enzymes.
That's done, again, in high-throughput by our test team. Our test team typically would lyse those organisms and provide the substrate for those enzymes and look for the desired products. They may, for example, be looking at, not just the activity of those enzymes, but also the specificity of those enzymes. It may be that you want one of the products of those enzymes but not another. Only at the end of that process do we select a specific amino acid sequence for the enzyme that we care about. We will then assemble that step of the pathway together with all of the other steps of the pathway for making the product we care about. Then at that point, we hand that over to our high-throughput fermentation team, whose job it is then to scale that pathway and that organism up for large scale industrial production.
So at each step there is application of robotic technology, all the way from building the pieces of DNA through the transformation, through the lysis of those, to the screening with things like LC-MS analysis of products, all the way through the robotics that's used for the fermentation scale up.
So how long does it take between that selection of the 1000 engineered enzymes that you're going to start to screen, to a final choice of what you will then scale up to be the final product?
The DNA assembly step takes probably four weeks, and the high-throughput test is probably another four weeks after that. A typical result of spending that time doing that testing will result in enzymes that are, I would say, on average probably about ten times better than the ones you started with. So this is a very worthwhile set of high-throughput experiments that really results in a significant improvement in the final pathway design.
Yeah, and it seems a very expedited process you're talking about, maybe a little over two months for a 10X improvement on an existing pathway.
Yes. Now, of course, what I didn't talk about is there's time at the beginning-
Trying to figure out what your pathway should be, and choosing the sequences, and there's a lot of time typically at the end involving scale up for fermentation, and also we've discovered if you're doing a commercial chemical product there's typically a large amount of resources and time spent in downstream purification of those final products.
Machine Learning and Predictive Biology
Absolutely. You mentioned that in the thinking before you begin the process of screening, say those 1000 compounds, you're doing a lot of data mining. So I wanted to get your thought on the future of predictive biology, and how those datasets are really going to play into changes that we see in biological research, really within the next decade or two?
Yeah. If you compare what we're currently doing in biology with other high technology disciplines, a good comparison would be the semiconductor industry, it's worthwhile thinking about how a company like Intel goes about producing their next generation microprocessor chip. What that process looks like is that their marketing team sets up a set of goals for their next processor, they assemble a team that might start out being as small as 10 or 20 people, and that team would start engineering the architecture of that microprocessor. That team would grow, probably to perhaps 1000 people over the next two years.
Those people would be involved in a very detailed design of that microprocessor, heavily, heavily relying on computer simulation to get every aspect of that processor correct. At the end of that two or maybe three year period, that team shrinks again, probably to 50 people, and at the end of that process they take the design of that processor, which is now thoroughly simulated, and they make it, and at the end of that fabrication process, they power it up, and unless they've really made serious mistakes it works the first time.
And there's no version two?
Well, there might be version 1.01.
There might be some slight tweak that they have to make, to make it perfect.
But basically there is no version two. We are very far away from doing that in biology.
And the question is, how do we transition from a technology of the kind that we have today, where we really are heavily reliant on experimentation. I hesitate to say it's blind, because it really isn't blind, but it's not driven by simulation and a detailed understanding of what's going on, but driven by an over reliance, in my view, on experimentation. So, how do we transition from where we are today to a predictive version of biology.
Yeah. And do we have the datasets in place to be able to do that?
No, absolutely we do not, and that really has been the major issue in trying to move forward in the field. It isn't just the datasets, it's actually in many cases a lack of knowledge about what's actually going on. I remember fondly some of my colleagues in the biology department at MIT, I would talk to them about wanted to engineer biology and they would say, some of them would say, "Tom, why are you working in E. coli? We already know everything there is to know about it, what could you possibly learn?" And there were a lot of things I didn't understand about how E. coli worked, and so I thought, "Ah, this is my opportunity. I will find out."
And what did you learn about E. coli?
I learned that they don't understand how it works either, and that was my takeaway from those discussions, which is that the scientific biology community, and of course this is a gross oversimplification, and perhaps even a misrepresentation, but many of them think that the relatively simple bacterial organisms are already so well understood that we have nothing left to learn from them, and that's absolutely not the case. We have huge gaping holes in our ability to think about what's going on in these organisms, and the only way of filling those holes is by measuring much, much more than we ever have before about what's going on inside those organisms.
One of the great resources that we have available at Ginkgo is that we are able to do very high-throughput measurements of a lot of these biological systems, and so my hope is that as we move forward over the next five or ten years, we will be able to apply some of those measurement technologies to build up the kind of dataset that's necessary to really start simulating these systems, rather than trying to just blindly engineer them. So that's a goal.
It's so wonderful to know, I think, that there is such a forward-thinking company that's really working towards building these datasets. At NEB we've been thinking a lot about artificial intelligence and machine learning and predictive biology. We've had some very interesting speakers from Google and other experts in the machine learning field, and it does seem like the data to drive these decisions is really where we're lacking in the biological fields right now. So it's so wonderful to know that there's a company out there working so hard to really build this resource that's going to benefit the world. So thank you for everything that you do.
Thank you for having me.
Yeah, thanks so much for being here today, it was such an honor to be able to interview you. I hope that you'll be enjoying the rest of your visit here today.
Thanks for joining us for this episode of our podcast. As always, check out the transcript of this podcast for links to further resources, and join us next episode for discussion of the history of molecular cloning, and speculations on where it's headed in the future, with New England Biolab's senior scientist, Bill Jack.
Explore NEB's other multimedia offerings:
|NEB TV Webinar Series
If you are enjoying our podcasts, you may also be interested in our NEB TV webinars.
Browse our extensive selection of videos by area of interest, and find helpful tips on product usage, troubleshooting guides, video protocols, scientific tutorials, and more.
View NEB's selection of brochures and newsletters in an interactive format.
About your host:
Lydia is a scientist by training and a communicator by nature, and has a knack for asking one too many questions.