Ginkgo Bioworks (YC S14) raises $100M to buy a lot of synthetic DNA
techcrunch.comI'm Jason Kelly, one of the Ginkgo co-founders -- happy to answer questions!
I'm a machine learning and have done work with Bayesian methods for modeling dynamic systems. I've been considering jumping into a synthetic biology company in industry. I've been searching for statistics and machine learning problems that need solving in synthetic biology. I suspect that there are two problems where a statistician or machine learning expert could contribute. The first is building data-driven models of metabolic pathways. The second is implementing an active learning approach to organism design -- basically building a robot that iteratively conducts experiments that maximize information while minimizing cost. But I also suspect that synthetic biology companies like yourself and Zymergen are more concerned with scaling up your business in the short term, and that implementing machine learning or computational biology-types of processes is a long term "optimization" task, not important to the core business in the near term. I'm afraid of making the jump if this type of work isn't important to the organization. Can you please comment? Cheers.
You mentioned two good examples of how a data scientist can contribute to a synthetic biology company. Models are useful in many ways but only if they are realistic and backed by data. Today, many models are limited in their usefulness because they make assumptions to reduce complexity, assumptions that are not always true in nature. We'd love to design iterative experiments and gather more data, so we can improve and expand these models. However, to end up with a useful outcome, we would need to a) test many designs and b) capture as many experimental parameters as possible. We are working on (a) -- through increasing ability to synthesize DNA, and by improving our foundry capability and scalability (so we can process and assay synthesized samples and analyze results). (b) is extremely challenging due to the complexity of biology. At Ginkgo, all data are analyzed by scientists and engineers with high degree of biological intuition, so they can fill in gaps not captured in data. For these reasons, we have focused our software and computation efforts on building up wetware and automation infrastructure, so we can run more and better experiments.
We are always looking for passionate engineers to join us to tackle tough challenges. Just because something isn't doable today doesn't mean we can't shoot for the moon! There's no better place to change how biology is engineered than here. Ping us if you are interested in joining our efforts.
Hi, I manage the data science team at Zymergen, and I can tell you we are working on these topics (and hiring). Shoot me an email if you'd like to chat -- dmitriy at zymergen.com
This is very cool stuff. I was just wondering what do you see is the biggest technical problem with this technology? I understand that the hardest part currently is transferring the modified genome into all cells, is this still correct?
If I'm interested in this technology how do you recommend learning the required techniques. As a machine learning engineer I know maths more then biology, but want to contribute to the open source movement. Where do you view the biggest impact of software/machine learning engineers can make to the open sourced biology movement?
The biggest technical problem in general for synthetic biology is predictive design. Most of the technology we are building is to enable massive numbers of designs to be tried in parallel b/c the complexity of the cell rebuffs predictive CAD models. This will change as we get more data which is what foundries like ours generate - lots of data.
I'd love to hear a reply to this.
Hello Jason Kelly and Patrick! I've been interested in synthetic biology for awhile now and am currently in the process of organizing a new iGEM team for my university. Did either of you compete in iGEM, and if so, how did it shape your perception of synthetic biology? Also, what advice would you give to a biochem student interested in pursuing SB as a career?
I believe Jason was on the first iGEM team! A lot of us at Ginkgo have been judges for iGEM, and many of our DNA Padawans are former iGEMers. iGEM is a great forum for synthetic biology because it encourages students to develop their research in creative and socially responsible ways.
In terms of SB as a career, the field is moving so fast it's better to focus on problem solving than techniques. Cloning will be obsolete soon; how would you design biological systems when you can synthesize any DNA sequence you want?
That's awesome! My school is small and doesn't have the same resources that some of the other schools do, but I'm going to push really hard to get a team started there.
I'm considering applying for the DNA Padawan internship after I graduate. Gingko Bioworks seems like a great place to work, and I bet I would learn techniques that are closer to the cutting edge than what I would learn in grad school.
I haven't tried my hand at SB yet (though I will soon!) but I suppose if I could synthesize any DNA sequence I wanted quickly and cheaply I would rely more heavily on computational intelligence approaches to come up with many solutions. Then I suppose you could print the DNA and test those solutions quickly, feed the data back to the algorithm and it would get better at designing future systems. (Computer science isn't my background, so I may have used some of those terms improperly).
Thanks for the reply randomscientist, you've made one (hopefully) future iGEMer really happy.
What's the current cost per base pair? Is a significant portion of the funding really going into synthesis?
Do you foresee synthesis costs falling quickly? It seems like its a major factor holding back commercial applications of synthetic biology.
Congrats on the funding and best of luck as you continue to scale!
This is Patrick, I lead our Design team here. Rather than commercial applications being held back by synthesis prices, it's the large scale demand from foundries ours that helps synthesis companies scale up their processes. If you assume that synthetic biology is going to be done by-hand for the foreseeable future, demand doesn't scale as quickly and prices stay high.
in other words, the robots are coming to steal our jobs?
Robots are coming to handle the tedious, manual labor, so that we can focus on more interesting pursuits. Then, the robots will be able to do that, and we can retire.
Hi, I'm interested in how much experience/prelim data you had and how large and experienced your team was when applying to ycombinator with this plan.
I applied with my partner with a similar goal a couple of years ago, and was rejected. The criteria for biotech applications was very vague then, and we tried to keep our pitch as non technical as possible which I think worked against us.
Also, do you have an in house gene design method or do you rely on the supplier to optimize? As I see it, the two major obstacles are reaching commercially viable expression yields, and identifying market demand.
> The startup announced in the spring of 2015 it would be buying 100 million base pairs of DNA to help it move into new production areas but has since upped that amount to the 600 million base pairs and will be partnering with Twist Bioscience and Gen9 to supply the synthetic DNA. Twist has pledged to deliver at least 400 of those base pairs of DNA by the end of 2017.
That last line about Twist delivering 400 base pairs has to be wrong, right? Do they mean 400 million?
Yes should be 400 million. 100 million from our last order plus 300 million in this new order.
Ah great, that makes a lot of sense. Thanks for the quick reply!
How do you compare Gen9 and Twist? Do you use each for a specific type of synthesis?
> works with DARPA to produce probiotics that will help U.S. soldiers stave off stomach bugs
Will there be synthetic probiotics in the future? (I understand that some probiotics are already heavily engineered.)