Biomedical Engineering Reference
In-Depth Information
Perfectly identical structures would have an RMSD of 0; matching short to moderate-length protein
structures typically have RMSDs in the 1-3 Angstrom range. A problem with RMSD is that it doesn't
take the size of the protein into account, and therefore the significance of the RMSD score can't be
taken as an absolute measure across all proteins. An RMSD of 5 or 6 Angstroms may be intolerable in
a molecule with only 50 residues, but perfectly acceptable in large protein molecules for applications
such as searching structure databases for known protein structures. However, even as a relative
measure, RMSD is valuable when working within a single family of proteins because the size of
structures will be about the same.
In addition to the RMSD measure, a variety of visualization techniques are available to provide
qualitative measures of similarity. Visualizing the protein structure is typically performed through the
use of any number of freely available protein rendering engines on the Web, such as RasMol,
Chimera, or SWISS-PDBViewer, as described in
Chapter 5
, "Data Visualization." An important note in
the use of visualization software as an adjunct to validation is that many visualization engines
incorporate embedded simulations in order to derive structure from simple bond angle data (see
Figure 9-19
) and may introduce uncertainty or bias in the output renderings. For example,
visualization packages that work with the PDB data format create images of molecules based on rules
and assumptions contained within the visualization package. Because rules and assumptions vary
from one tool to another, there may be significant differences among the graphic output produced by
these systems. In contrast, tools expecting ANS.1 format data don't rely on an embedded simulation
to create the visual output, and their output is free of bias.
Figure 9-19. Embedded Simulation. Many visualization tools incorporate
embedded simulation systems (top) in order to create the data necessary
for visualization. Tools that expect data in the PDB data format are included
in this category.
