Calculation of The Protein Structure Quality Score (PSQS)
PSQS is calculated in two steps:
Geometrical parameters of a given structure are obtained including: local backbone
geometry class, burial status of all residues (based on the number of interatomic
contacts with other residues), and contact distances between different residues
Calculation of local, burial and contact energy is accomplished by application
of statistical potentials of mean force to these parameters
We have found that in many cases PSQS makes it possible to differentiate
between correct and incorrect protein structures. It should be noted, however, that
essentially PSQS reflects the similarity between a given protein structure and
a "typical protein structure", so it is expected to yield best results for monomeric
globular proteins without large ligands or significant biases in amino-acid
PSQS may be used in homology modelling to evaluate alternative protein models
based on different alignments. It may also be useful during structure refinement
to locate the places that are significantly different from typical geometries.
The graph on Fig. 1 shows the joint empirical distribution density of psqs and rmsd.
The distribution was calculated on the data base of 41,000
Levitt Decoys. As expected PSQS values increase
with increasing distance from the native structure.
This relationship holds within about 5 A RMSD from the native structure.
Fig. 2 shows the distribution of PSQS on a representative set of structures covering all folds taken
from the SCOP data base. The average PSQS for this set is -0.27 and most have structures have
PSQS less than -0.1. These statisics were used to calibrate the coloring scheme used for analysis of
3D distribution of PSQS along the protein chain.
Figure 1 (the graph was made with gnuplot)
Jaroszewski L., Pawlowski K, Godzik A, Multiple Model Approach: Exploring
the Limits of Comparative Modeling, J.Mol.Model. 1998, 4, 294 - 309
K, Jaroszewski L, Bierzynski A, Godzik A, Multiple model approach--dealing
with alignment ambiguities in protein modeling,Pac Symp. Biocomput. 1997,
Godzik A, Kolinski A, Skolnick J, Are proteins
ideal mixtures of amino acids? Analysis of energy parameter sets. Protein
Sci 1995 Oct;4(10):2107-2117