In this interview, Zak Costello, Machine Learning Senior Scientist, shares his perspective on the unique strengths and challenges facing Generate:Biomedicines in the rapidly evolving field of generative biology.
Before joining Generate, Zak specialized in control theory, a branch of engineering and mathematics focused on developing models for controlling systems through feedback – like how a thermostat keeps a room at just the right temperature. Searching for an area of biology to which he could apply this kind of thinking, he dove into the challenge of engineering protein sequences and structures. His work in this area brought him to the attention of a stealth startup that eventually became Generate, and he’s been a key member of the Machine Learning team for nearly five years.
You’ve been at Generate for almost five years now. What has been the most meaningful part of your work here? Why does what you do matter to you?
There are two parts for me. One is I find real joy working with, learning from, and mentoring incredibly talented people who are oriented around scientific discovery and have an underlying desire to use that to help people in some way. I find that extremely gratifying.
The other thing that is very meaningful to me is that I truly believe in the direction of the company. When you think about it, what we’re doing is very strange. We are trying to understand how humans are built – how life is constructed – to learn principles that we can use to create machines that can modify those things. It’s incredibly powerful to be learning about and contributing to what is, at its core, a very ancient process.
What is one major challenge in generative biology, and how are we working to overcome this issue?
In terms of the harder technical challenges, the problem at the forefront is protein dynamics and understanding how proteins interact. For example, we’re working to understand what allows one protein to bind and connect with some things but not others. That’s a major focus of what we’re working on now, and it will be a cool advantage for us if we figure it out. It’s going to open the door to a lot of potential opportunities down the road.
However, I don’t think the technical challenges are our biggest obstacles. We certainly have a lot to figure out about generative biology, but the real challenge is getting people to work together and understand the biology, the proteins, and the specific tasks we want to use them for. Our ability to integrate the talents of colleagues with computational science, biologic engineering and clinical expertise will be the thing that really leads to breakthroughs.
What advantages does Generate have in our space?
There are so many different kinds of scientists and experts here, all working together on big problems. I think that’s what allows us to really solve these problems at scale. We have collectively internalized that to be successful at protein engineering, we need a full spectrum of expertise with people working on every part of the process. I think that knowledge comes from a deep appreciation for the complexity and difficulty of biology itself.
Engineers (and physicists) work on complex, technical problems, and they work hard to develop deep expertise in one or more areas. Often these areas have natural laws that are reliable and closely model reality—there aren’t a lot of exceptions to the rules. They can rely on the tools they have to build great things. When people from these backgrounds move into biology, where hard and fast rules are rare, they might think: “I’ve succeeded with these powerful methods in other areas. I bet I can solve biology’s longstanding problems – like protein design – in a similar way.” But most of the time when these attempts are made, they will find, again and again, that while their models work in theory, they don’t work in the lab.
Biology is incredibly deep, complex, and messy. Over time humility emerges from failure, and hard-won successes in biology are better contextualized and appreciated. No one person or expert has the whole picture. And I think Generate has done a good job from the very beginning of valuing different kinds of expertise that can help give us the whole picture as a company.
What is one piece of advice from your own career (or your life in general) that you found very valuable?
One value that was modeled to me by mentors throughout my career is the importance of intellectual openness. As a scientist, I try to be very careful with how attached I get to my own ideas, and instead I try to be open to criticism, supportive of new ideas that might lead to better collaborations, and comfortable with a certain rate of failure.