 |
Ridgway
Scott makes computer-screen stars of molecules, money, and more.
The mathematician and computer scientist hones the high-tech tools
of
biomedical, financial, and other researchers.
When a computer-software program crashes, others might curse their
screens and huff away. But that’s just the moment Ridgway Scott
wants to pull up his chair. Amathematician and computer scientist,
Scott works on some of the “buggiest”—or most problematic—software,
a challenge he says he enjoys because “the most interesting
mathematics come out of this process.”
The math he’s talking about goes far beyond the use of numerals
in the binary code that most computer languages are written in.
He means the nitty-gritty—the underlying differential equations
and numerical algorithms—that run not simple word-processing
programs, like the ubiquitous Microsoft Word, but that can analyze
large amounts of data, from genetic information to stock-market
quotes, or simulate on a computer screen the real-life behavior
of something very, very small or very, very big, whether an enzyme
or a star. And his preferred PC is not a personal computer but a
parallel computer, which has the power to harness thousands of computer
processors and perform billions of computations at once.
A former L. E. Dickson instructor in mathematics at the U of C,
Scott returned to Chicago last fall as a professor in the departments
of computer science and mathematics. He earned his Ph.D. in mathematics
in 1973 from MIT, where he was a pioneer in refining the finite-element
method, the most widely used computational technique for engineering
design and analysis. He later helped to establish parallel-computing
centers at Pennsylvania State University and the University of Michigan.
Most recently, at the University of Houston, he directed the Texas
Center for Advanced Molecular Computation, a research group devoted
to biomolecular design and funded by the National Science Foundation.
He has summed up his findings in two books and more than 100 papers
on structural mechanics, fluid dynamics, nuclear engineering, computational
chemistry, and daunting mathematical techniques—with names
like “boundary element,” “finite difference,”
and “spectral”—that are used to solve the partial-differential
equations applied in engineering.
Although he’s still moving into his new office in the Ryerson
Physical Laboratory building, Scott has long since unpacked one
box: the box that contained his custom-built desktop computer. Its
enviable specifications include two Intel Pentium II processors,
a half gigabyte of RAM, and a 10-gigabyte hard drive; moreover,
it’s linked to several others—nearby and just like it—to
form a parallel computer. At Chicago, Scott’s using the high-powered
machine to continue his previous studies and to launch some new
software projects that will aid geneticists, financial analysts,
and even astrophysicists.
In November, Daphne Preuss, an assistant professor in molecular
genetics & cell biology, asked Scott to help her manage data
collected as part of her study of plant genetics. She had been using
her word-processing program to pinpoint repetitive DNA sequences,
a painstaking process. Scott’s now looking at ways the sequences
can be graphically presented. He’s also collaborating with
senior lecturer Robert Almgren to create software that lets students
in the master’s program in financial mathematics model stock-market
scenarios using massive amounts of New York Stock Exchange data.
And at the University/Argonne Center on Astrophysical Thermonuclear
Flashes, he is helping to develop software that can simulate the
violent explosions occurring when hydrogen from one star accumulates
on another nearby star and ignites.
These additional projects have not subtracted from Scott’s
ongoing work. He’s expanding his efforts in the field of computational
fluid dynamics through a new research group being organized at Chicago
by physicist and mathematician Leo Kadanoff. Scott’s work in
this area might inform, for example, engineers designing steel-making
systems that are based on complicated geometrical patterns.
He’s also producing results as the project leader of two research
teams organized through the National Partnership for Advanced Computational
Infrastructure (NPACI), a group of 46 research institutions and
universities exploring how the computational power of parallel computers
can most easily be applied in science and engineering.
One of Scott’s teams is refining computer-generated images
of molecules. The team has already conducted simulations that reveal
an open “side door” in the enzyme acetylcholinesterase,
or AChE—a finding that may aid in the making of pharmaceutical
drugs that target AChE. While clinical studies suggest, for example,
that AChE inhibitors may be useful in enhancing memory in patients
with Alzheimer’s disease, an effective inhibitor cannot be
designed without a detailed understanding of the AChE molecule and
how it might interact with the inhibitor.
That’s where Scott and his team enter the equation. It’s
their job to figure out how to model the behavior of the more than
130,000 atoms involved in such a show. They have pushed the limits
of the national supercomputer center at the University of California,
San Diego, requiring no less than 128 processors to conduct one
simulation.
The work of Scott’s other NPACI team may ease this process
in the future. In February it released a new set of computer languages,
called the P-languages, which, explains Scott, can reproduce structures
with irregular shapes and random movements—like molecules—better
than previously used languages that work best when applied to predictable
grid-like patterns.
It’s Scott’s transferable skills that move him so easily
from the cosmic to the atomic level and everywhere in between. He
breaks down his approach to developing scientific-simulation software
into these basic stages: Donning his mathematician’s hat, Scott
represents the force of electrical charges and the other known physical
laws affecting a particular molecule, for example, as differential
equations—some already devised, others he must create. Next,
he derives numerical algorithms to solve the equations. The computer
scientist Scott then translates the algorithms into a computer code
before his applied-math side uses the resulting simulation program
to study the molecule’s behavior.
“The typical scientist or engineer just wants to load the program
and run it,” he says. “I like that too, but it’s
the process leading up to the code development that I find most
interesting. If we don’t do our job, it ain’t going to
run.”
Scott hopes to provide an “intellectual home” for all
the software experts on campus who, like him, apply their skills
across disciplines. He’s working with Robert Zimmer, deputy
provost for research, and Rick Stevens, director of the mathematics
and computer science division at Argonne, to form a computation
institute that would foster the interdisciplinary development of
software for use in the biological, financial, physical, and social
sciences as well as the arts and humanities.
“There are lots of people in areas as diverse as linguistics,
biology, and physics at the forefront of the computations field,”
says Scott. “The institute should foster synergy among like-minded
people in different departments.”—C.S.
|
