Homology Modeling: Generating Structural Models to Understand Protein Function and Mechanism - Computational Modeling of Biological Systems - page 108

Fig. 6 Steric clashes in homology models. Homology models on an average feature much higher
extent of steric clashes when compared to experimental structures ( a ). The distribution of clash-
scores (which is a normalized energetic parameter reflecting the extent of steric clashes in a protein
structure [ 50 ]) of high-resolution crystal structures and representative homology models from
SWISS-MODEL database are plotted. The efficacy of Chiron in minimizing clashes in protein
structure is demonstrated for the homology model of Q13823, whose initial model ( b )hasa
clash-ratio of 0.13, much higher than that seen in experimental structures. The protein structure
is shown with the cartoon representation, rendered using PyMol ( http://www.pymol.org ). Clashes
are denoted as colored cylinders , where both the colors and the thickness of the cylinders denote
the van der Waals repulsion energy. The scale of the repulsion energy is shown as a gradient bar
at the bottom , with the numbers at the ends indicating repulsion energy in kcal/mol. Note the
large numbers of cylinders in the initial model, denoting excessive steric clashes. The minimized
structure ( c ) has a clash-ratio of 0.018, within one standard deviation of the mean clash-score of
high-resolution structures
The refined structural models by definition should feature physically reasonable
backbone conformation and a well-packed core that has an acceptable extent of
clashes. Minimal backbone perturbation to ensure ideal packing can be achieved
by various means including “backrub” and knowledge-based backbone assembly
(as used in Rosetta), all-atom DMD simulations and minimization using molecular
mechanics forcefields. All these methods refine the structural model to the nearest
local minima in the conformational space of the starting structure. Thus, if the