STRUCTURES BY COMPUTATION
Software predicts structure and binding of membrane proteins
Starting with nothing but a string of amino acids, a new set of computer programs appears to be able to predict the structure and binding characteristics of membrane proteins.
The ability to make such accurate predictions has been the goal of many scientists for decades, and it would be an enormous boon to a range of researchers, including drug designers and enzymologists. William A. Goddard III, director of the California Institute of Technology's Materials & Process Simulation Center, and colleagues report that the methods they've developed have predicted the previously unknown structures of several important proteins [Proc. Nat. Acad. Sci. USA, 99, 12622 (2002)].
CALCULATED Epinephrine nestles in predicted binding site of predicted human 1-adrenergic receptor.
The proteins they chose to model, known as G-protein-coupled receptors, or GPCRs, are ubiquitous membrane proteins that play a role in processes such as vision, pain, allergies, and brain diseases. Molecules that bind to the proteins' receptors on the outside of the membrane activate so-called G proteins on the other side, which in turn precipitate cascades of biological processes.
But membrane proteins in general are so difficult to crystallize--they tend to denature in aqueous solvents--that a 3-D X-ray structure exists for only one GPCR, bovine rhodopsin.
In addition to bovine rhodop-sin, Goddard's group studied a variety of GPCRs, including the human sweet receptor, the -adrenergic receptor, and mouse and rat olfactory receptors. The structure prediction program, called MembStruk, calculates the complex shape assumed by proteins, starting only from the amino acid sequence.
The predicted bovine rho-dopsin structure compares well with crystallographic data, the group reports. The protein structures for which no experimental data exist, however, were trickier to validate. Using a program they developed called HierDock, Goddard's group identified ligand binding sites on the proteins, then docked into them dozens of molecules for which binding data already exist. The predicted binding energies and conformations also agree well with the experimental data, particularly so with the -adrenergic receptor, they say.
The work "represents an outstanding contribution toward addressing the structure prediction and the function prediction of G-protein-coupled receptors," says Christodoulos A. Floudas, chemical engineering professor at Princeton University.
"If it applies generally to the prediction of other membrane proteins and other GPCRs, [this work] could have an enormous impact on pharmaceutical de-sign, since many of the current blockbuster drugs interact with membrane proteins," says Ken A. Dill, professor of pharmaceutical chemistry at the University of California, San Francisco, an expert in computational protein structure prediction.
Goddard tells C&EN, "I believe that we are at the point where the 20 years of struggling with protein-folding predictions are finally paying off."