Comparative Protein Structure Modeling.

Three-dimensional protein structures are invaluable sources of information for the functional annotation of protein molecules. These structures are best determined by experimental methods such as X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy. However, the experimental methods cannot always be applied. In such cases, prediction of the protein structure by computational methods can frequently result in a useful model. Protein structures can be modeled either ab initio from sequence alone or by comparative methods that rely on a database of known protein structures (1,2). Ab initio methods are largely based on the laws of physics, while comparative methods, including comparative (or homology) modeling and threading, are based primarily on statistical learning. Although there have been significant improvements in the ab initio (3) and threading methods (4), comparative modeling gives the most accurate results if a known protein structure that is sufficiently similar to the modeled sequence is available (1). [ABSTRACT FROM AUTHOR]