Winter 2008

Living Science

As a mathematician, Julie Mitchell was first drawn to proteins because of their interesting geometry. Now, she uses math as a tool to study protein function.


Julie Mitchell, an assistant professor of biochemistry and mathematics, builds computational tools that help biologists analyze and predict changes to protein structures that will alter their function—information that can help scientists design novel drugs and disease-fighting agents. Describing herself as a liaison between the logic-driven world of math and the chaotic realm of biology, Mitchell has collaborated with researchers working on everything from cancer to better processes for making ethanol.

How different are mathematics and biology anyway?

In many ways, mathematics and biology are almost logical opposites. Mathematics is more of an art: You create rules for some geometric or algebraic space, and then you see what you can prove about it. In biology, we can’t even figure out what some of the basic rules are. There are no theorems. Everything has a little asterisk that says: except in these following cases. And so, as a mathematician, if you want to go into biology, you have to be able to accept that and not allow it to drive you insane.

Is that how you came into this field, as a mathematician?

Yes, I graduated with my math degree in 1998 from UC–Berkeley.

What was it that attracted you to biology and proteins?

My background is in geometric analysis, so what drew me to proteins in the first place was their interesting shapes. Some people look at proteins mainly as strings of amino acids, and when they think about what we call a point mutation—a change to a single amino acid—to them that’s a change in the linear sequence. I don’t quite think that way. To me, that’s a structural change, a geometric change, as well as a change in the protein’s biochemical properties.

Now for a long time, I think I translated chemistry into mathematics through physics. You know, everything is a bunch of atoms, here are the forces on the atoms, and so on. That can be a starting point, but ultimately you have to begin thinking a bit more like a biochemist—which takes a lot of time.

How did you begin to do that, to think more like a biochemist?

(laughing) I think trial by fire. When I first switched fields, the most helpful thing I did was not to take classes or read books, but to attend a lot of professional meetings and go to the talks and, especially, the poster sessions. That’s where the graduate students and postdocs are representing their papers, and you feel a little less intimidated saying, “I know this is a really dumb question, but what does that mean?”

Not being afraid to ask a dumb question is very important. You have to be very brave and humble, in a way, to admit when you don’t know something. At the same time, you have to be very good at—how shall I say this?—pretending to know what you’re talking about when you half-don’t (laughing).

It truly does sound like trial by fire. Why go through it?

When I was finishing up my Ph.D., I really loved mathematics, but I didn’t feel a tangible connection between the work I was doing and the real world. Mathematics was beautiful, it was fascinating, but I felt like I wanted to have a more immediate impact.

Eventually, I was pointed toward the protein-folding problem, and so I got on the Internet and typed in “protein folding.” And then I pulled up pictures of these fabulous proteins, with all of these interesting coiled structures, all wrapped up into a compact little ball. And there were partial differential equations and dynamical systems that went along with these fascinating geometric structures, and from that point on, that’s what I was doing.

Did you have any grasp of the biological significance of proteins back then?

I had no clue. Honestly, before I started on this quest, I didn’t know there was more than one kind of protein, such as antibodies or enzymes. Protein was something on the back of the box in the grocery store—it was this generic, singular thing. Since then, I’ve come to understand the diversity of the world of proteins and all the important functions they perform.

So what can math reveal about proteins?

Mathematics can be many things, but one of the fundamental capabilities of math lies in recognizing and characterizing patterns. Within the realm of biology, math and statistics can help uncover patterns in biological systems, like proteins, that aren’t immediately obvious, or may only be obvious to an expert.

For example, the predictions made by our computational tool, called the KFC server, are certainly possible for a trained structural biologist to make, but the analysis might take hours rather than seconds, and it can’t be automated. Computational mathematics makes the prediction process faster, more statistically non-specialist.

What specifically have you been trying to understand about proteins?

Well, there are many different questions we’ve posed. One of them is protein docking. Given two structures of proteins, can you predict how they come together? It’s a bit like solving a three-dimensional puzzle, only a lot harder.

Now, the KFC server solves a somewhat different problem. If we already know how two proteins come together, can we identify, among all the amino acids in the interface between them, which ones are most important in the sense that if you change them, the proteins won’t bind as well anymore? Then there’s a somewhat harder problem, which is can you predict mutations that are going to improve the binding?

Fundamentally, we want to provide tools to experimental groups, particularly to people who aren’t necessarily used to computational tools or who don’t usually give a lot of thought to protein structure. How can we help them identify the structurally important features of their protein systems?

Is that why you’ve chosen to offer your tools on a server over the Web, because you’re targeting non-mathematicians?

Yes, the beauty of doing things with a server is that nobody has to download code, or run an application. Everybody, including your grandma, can work the Internet. Well, maybe most people’s grandmas.

My philosophy is that you should be able to run a problem using very few parameters. You also don’t want people to waste a lot of time typing in commands when you can automate things for them with a check box or a drop down menu, or something else that’s going to allow them to very quickly get at the information they need. That’s really the goal of the KFC server: to make it very simple for people to interact with protein-protein interfaces and to highlight the important features of those interfaces in a way that’s intuitive and simple to use.

Where do you see yourself going from here?

Since I’ve come here, I’ve become interested in bioenergy applications. The most important problem in the world right now—by far the most important problem—is climate change and energy.

I feel that in the last decade there has been so much emphasis on medical science and medical advances—it’s fabulous. On the other hand, if you increase everybody’s life span by 20 percent—just to throw a number out there—you’re increasing the impact of each individual on the planet by 20 percent. Therefore, you have to balance medical advancements with advances in agriculture and energy science, and so on. So to be able to have any impact on that at all, is just … it would be an honor, if that makes sense.

So can math save the planet?

(laughing) Well, we’ll do our best.

This article was posted in Environment, Features, Health, Living Science, Winter 2008 and tagged , .