Communications - October 2009

nancial world applying quantitative and computational techniques to the process of investment management. During the early years of D. E. Shaw & Co., the financial firm, I’d been personally involved in research aimed at understanding various financial markets and phenomena from a mathematical viewpoint. But as the years went by and the company grew, I had to spend more time on general management, and I could feel myself getting stupider with each passing year. I didn’t like that, so I started solving little theoretical problems at night just for fun—things I could tackle on my own, since I no longer had a research group like I did when I was on the faculty at Columbia. As time went by, I realized that I was enjoying that more and more, and that I missed doing research on a full-time basis.

I had a friend, Rich Friesner, who was a chemistry professor at Columbia, and he was working on problems like protein folding and protein dynamics, among other things. Rich is a computational chemist, so a lot of what he did involved algorithms, but his training was in chemistry rather than computer science, so when we got together socially, we often talked about some of the intersections between our fields. He’d say, “You know, we have this problem: the inner loop in this code does such-and-such, and we can’t do the studies we want to do because it’s too slow. Do you have any ideas?”

Although I didn’t understand much at that point about the chemistry and biology involved, I’d sometimes take the problem home, work on it a little bit, and try to come up with a solution that would speed up his code. In some cases, the problem would turn out to be something that any computer scientist with a decent background in algorithms would have been able to solve. After thinking about it for a little while, you’d say, “Oh, that’s really just a special case of this known problem, with the following wrinkle.”

One time I managed to speed up the whole computation by about a factor of 100, which was very satisfying to me. It didn’t require any brilliant insight; it was just a matter of bringing a bit of computer science into a research area where there hadn’t yet been all that much of it.

At a certain point, when I was approaching my 50th birthday, I felt like it was a natural time to think about what I wanted to do over the coming years. Since my graduate work at Stanford and my research at Columbia were focused in part on parallel architectures and algorithms, one of the things I spent some time thinking about was whether there might be some way to apply those sorts of technologies to one of the areas Rich had been teaching me about. After a fair amount of reading and talking to people, I found one application—the simulation of molecular dynamics—where it seemed like a massive increase in speed could, in principle, have a major impact on our understanding of biological processes at the molecular level.

It wasn’t immediately clear to me whether it would actually be possible to get that sort of speed up, but it smelled like the sort of problem where there might be some nonobvious way to make it happen. The time seemed ripe from a technological viewpoint, and I just couldn’t resist the impulse to see if it could be done. At that point, I started working seriously on the problem, and I found that I loved being involved again in hands-on research. That was eight years ago, and I still feel the same way.

hAnrAhAn: In terms of your goals at D. E. Shaw Research, are they particularly oriented toward computational chemistry or is there a broader mission?

shAW: The problems I’m most interested in have a biochemical or biophysical focus. There are lots of other aspects of computational chemistry that are interesting and important— nanostructures and materials science and things like that—but the applications that really drive me are biological, especially things that might lead not only to fundamental insights into molecular biological processes, but also to tools that someone might use at some point to develop lifesaving drugs more effectively.

Our particular focus is on the structure and function of biological molecules at an atomic level of detail, and not so much at the level of systems biology, where you try to identify and understand networks of interacting proteins, figure out how genetic variations

affect an individual’s susceptibility to
various human diseases, and so forth.
There are a lot of computer scientists
working in the area that’s commonly
referred to as bioinformatics, but not
nearly as many who work on prob-
lems involving the three-dimensional
structures and structural changes that
underlie the physical behavior of in-
dividual biological molecules. I think
there’s still a lot of juicy, low-hanging
fruit in this area, and maybe even some
important unifying principles that
haven’t yet been discovered.
hAnrAhAn: When you mention drug
discovery, do you see certain applica-
tions like that in the near-term or are
you mostly trying to do pure research at
this point?
shAW: Although my long-term hope
is that at least some of the things we
discover might someday play a role in
curing people, that’s not something
I expect to happen overnight. Impor-
tant work is being done at all stages
in the drug development pipeline, but
our own focus is on basic scientific re-
search with a relatively long time hori-
zon, but a large potential payoff. To put
this in perspective, many of the medi-
cations we use today were discovered
more or less by accident, or through
a brute-force process that’s not based
on a detailed understanding of what’s
going on at the molecular level. In
many areas, these approaches seem to
be running out of steam, which is lead-
ing researchers to focus more on tar-
geting drugs toward specific proteins
and other biological macromolecules
based on an atomic-level understand-
ing of the structure and behavior of
those targets.
The techniques and technologies
we’ve been working on are providing
new tools for understanding the biol-
ogy and chemistry of pharmaceutically
relevant molecular systems. Although
developing a new drug can take as long
as 15 years, our scientific progress is
occurring over a much shorter tim-
escale, and we’re already discovering
things that we hope might someday be
useful in the process of drug design.
But I also enjoy being involved in the
unraveling of biological mysteries,
some of which have puzzled research-
ers for 40 or 50 years.