Freely distributed models could help speed discovery
of new drugs
David Terraso
Institute Communications and Public Affairs
In an important step toward accelerating drug discovery,
researchers have created computer models of more than
900 cell receptors from a class of proteins known to
be important drug targets. The models, which are now
freely available to noncommercial users, promise to
help scientists narrow their research inquiries, potentially
speeding up the discovery of new drug compounds. The
research appears in the February 17 issue of the Public
Library of Science Computational Biology.
“This is the first time anyone has modeled them
all with an algorithm that improves the accuracy of
the structure,” said Jeffrey Skolnick, Georgia
Research Alliance Eminent Scholar in Computational Systems
Biology. “I think it’s going to have significant
impact, because it’s a major class of drug design.”
One of the hottest areas in drug research, rational
drug design uses three-dimensional computer
simulations to study how different drugs and their cellular
targets interact with each other. This technique can
help research teams discover which compounds are most
likely to achieve the desired results, potentially accelerating
the speed of drug research and allowing for the discovery
of reactions that may not have been found through traditional
means.
The mission of Tech’s Center for the Study of
Systems Biology, of which Skolnick is the director,
is to essentially simulate life on a computer by building
accurate three-dimensional models of the components
of life, such as individual proteins and collections
of proteins.
“The idea is to simulate these proteins, introduce
a drug structure and see how they interact,” he
said.
G protein-coupled receptors (GPCRs) are targeted by
an estimated one-third of all drugs and convey chemical
signals from the outside of cells to the inside. But
since they tend to fall apart once they’re removed
from the outer membrane of the cell, scientists have
only been able to solve the three-dimensional structure
for a few of them.
And those aren’t even good drug targets. Until
now, researchers wanting to model any of the others
have had to base their models on the structures of the
existing, non-pharmacological receptors. According to
Skolnick, the models aren’t very accurate because
those receptors are evolutionarily distant from the
proteins thought to be good drug targets.
Using an algorithm they developed known as TASSER,
a team of researchers led by Skolnick created three-dimensional
structures of all the GPCRs below 500 amino acids in
the human genome.
“The solved GPCRs are of the same approximate
shape as the ones known to be good drug targets, only
they differ in details. But it’s the details —
the packing of the helixes, their angles, their size
— that differentiate the drug binding sites of
GPCRs from one another,” said Skolnick, “TASSER
appears to have the capacity to give us a reasonable
picture of the structure of these proteins.”
Of the 907 models TASSER has helped create, Skolnick
estimates that about 820 are accurate enough to be useful
to researchers.
“There’s still room for significant improvement.
They’re like cartoons — they kind of look
like reality sometimes, but they can be used to help
design experiments,” said Skolnick.
The next step for Skolnick is solving the structure
of proteins that have been implicated as a factor in
various types of cancer.
| 
