Biology Reference
In-Depth Information
one of the most frequently studied classes of drug targets. These include
protein families and superfamilies such as ion channels, reuptake pumps
as targets for antidepressants, and the important group of seven-trans-
membrane G-protein-coupled receptors (GPCRs). However, membrane
proteins pose formidable challenges to experimental structure determi-
nation by X-ray crystallography and NMR spectroscopy. Furthermore,
human proteins often have no closely related homologs in prokaryotes or
archaea, which would facilitate expression and crystallization. As a result,
structures of membrane proteins are significantly underrepresented in the
PDB. The 3D structures of only
160 different membrane proteins are
currently publicly available (June 2008). Consequently, prediction of
membrane protein structures based on physical models that describe
intraprotein and protein-solvent interactions in the membrane environ-
ment without relying on homologous template structures has been
attempted by several groups.
55,56
An important challenge in the modeling of membrane protein struc-
tures is the presumed difference relative to globular proteins. For exam-
ple, it is believed that membrane proteins are “inside-out” globular
proteins, with hydrophobic residues on the outside in contact with the
lipid bilayer and polar residues on the inside in the protein core. This
design may render the standard scoring functions used for the modeling
of globular proteins less suitable for use with membrane proteins.
Recently, a new scoring function was developed in Rosetta to account for
such differences.
57
∼
4.3. Model Quality Evaluation
Protein structure modeling is maturing and therefore widely used as a
scientific research tool today. Consequently, it is increasingly important
to evaluate to what extent current prediction methods meet the accuracy
requirements of different scientific applications. A good way to assess the
reliability of different protein structure modeling methods
a posteriori
is
by evaluating the results of blind predictions after the corresponding pro-
tein structures have been determined experimentally. One such effort is
the biannual Community Wide Experiment on the Critical Assessment of
Techniques for Protein Structure Prediction (CASP).
7,58
During a CASP
