Jerome P. Nilmeier, a
biophysicist working in computational biology, is willing to bet his new
research will provide a breakthrough in the use of the Monte
Carlo probability code in biological simulations.
Working with Gavin E.
Crooks at Lawrence Berkeley National Laboratory, David D. L. Minh at Argonne
National Laboratory, and John D. Chodera, from the University
of California, Berkeley,
Nilmeier has coauthored a paper that introduces a new class of Monte Carlo moves based on nonequilibrium dynamics. The
paper appears in the Proceedings of the National Academy
of Sciences.
The Monte
Carlo technique is one of the most widely used methods to model a
system and determine the odds for a variety of different outcomes. The
technique was first developed by scientists working on the Manhattan Project
who needed to figure out how far neutrons might pass through a variety of
different types of shielding materials.
The Monte
Carlo technique harnesses the power of computers to figure out the
probable outcomes of equations that have hundreds or thousands of variables. It
is a short cut that, instead of giving a definitive answer, gives a probable
answer. Random numbers are put into the equation, the outcome is tested, the
probability for the different outcomes is determined and then a decision can be
made about what is the most likely outcome.
To help explain the
impact of the new Monte Carlo technique,
Nilmeier uses the example of a chemical compound composed to two identical or
similar subunits, what’s known as a dimer.
“The dimer model
that we studied is a good proxy for a reactive system where molecules are
allowed to collide and form new molecules and also to dissociate into free
atoms,” Nilmeier says. “Using the new technique, we can bias our
simulation to sample the collision event more frequently and obtain better
statistics. In our paper, we report an order of magnitude improvement, but we
can easily imagine systems that would have several orders of magnitude in
improvement.”
One of the important
keys to the Monte Carlo technique is how it
goes about making guesses. Different types of systems have different types of
variables with different types of relationships. What works well for measuring
how far a neutron will pass through different type of radiation shielding
doesn’t work well at all when it’s applied to a biological system where there
are millions of molecules all moving very short distances very rapidly in many
different directions. What Nilmeier and his colleagues figured out is a new
method to apply the Monte Carlo technique to
these types of problems.
“My colleagues and
I were pointed to this new method in 2008. We were all at the annual Berkeley
Mini Stat Mech Meeting when we noticed new theorems that could be applied to
biology,” says Nilmeier. “Following the meeting, we started to
work.”
Nilmeier currently works
on a team led by Carol E. Zhou of the Global Security Computing Applications
Division and Felice Lightstone, the group leader of the biosciences and
biotechnology division.
