Any interesting medical protein folded with a complete primary to quaternary structure will act as a catalyst to drive a
biochemical reaction. These reactions, which are the basis of all biological function, involve sites on the surface of
the protein that bind to a particular ligand or macromolecule. However, these reactions are not always favorable, and
sometimes there is a need to control their rate or stop them completely. A competitive inhibitor, another ligand of
similar structure to the original, can slow down the reaction by diluting the original ligand and decreasing the probability
of it occupying the site. However, this direct method will only slow down the reaction. To stop the reaction, another
solution is needed.
The solution to this problem is to not deal with the original site itself, but instead use an indirect approach. By finding
another site that will cause a change in the folding of the protein, it is possible to disable the ability for the original
site to react. Such a site is called an allosteric site, and when occupied, an allosteric active site deforms or eliminates
the original site in what is referred to as allosteric inhibition. Though simple conceptually, this method presents quite
an engineering challenge. Allosteric sites exist on the surface of proteins similar to the original reaction sites.
However, unlike the original site, allosteric sites do not always have ligands in the system that fit them. This makes
it difficult to determine the sort of molecule structure needed to activate the allosteric site.
The current biochemical method of allosteric site discovery involves an analysis of the protein structure with sequencing
and x-ray crystallography, followed by a guess-and-check chemical testing method for finding possible drugs. From an
engineering standpoint, this method is not useful in that it cannot be automated efficiently. Instead, our solution is
to use the protein structural data given by X-Ray crystallography to compute the locations and properties of these allosteric
sites. This can in turn derive the properties of ligands that could activate them.
The work we have done on this project over the past two years has already yielded an algorithm named CIAM (Candidate
Identification Allosteric Modeling) that accomplishes some of these goals. By analyzing the surface of the protein structure
and taking other factors into account such as domain mobility, CIAM can already generate features to help locate possible
allosteric sites and their properties. This project proposes a two part extension to the CIAM functionality. First, by
integrating a molecular dynamics package into the CIAM algorithm, it is possible to eliminate the last bit of human busy work
required to process a particular protein. Second, by developing an algorithm to find and rank these allosteric sites, we can
use the CIAM data to accurately map potential allosteric sites and model the types of ligands that could activate them. Using
a customized graphical interface, the biochemist can then review the results and chemically test ligands with matching properties
for the most highly ranked sites.