Homology Modeling: Generating Structural Models to Understand Protein Function and Mechanism - Computational Modeling of Biological Systems - page 111

The standard values of bond lengths and angles are obtained from studies on
model small molecules [ 61 ] or through surveys of high-resolution crystal structures
[ 62 ]. The backbone torsions comprise of the '; ,and! angles. The ' of
each residue can be compared to the allowed ' map obtained from survey
of high-resolution crystal structures to detect outliers. The side-chain torsions
are again compared to the rotamer libraries [ 63 ] to detect outliers. All these
measures ensure that the covalent geometry of the structural model is physically
acceptable. The packing quality can be assessed based on the extent of steric clashes
[ 50 ], the prevalence of voids, and the scaling of the solvent accessible surface
area with protein length [ 62 ]. There are several servers that can compare these
structural parameters of a model with benchmark distribution to indicate the areas
of the protein structure that need further refinement to be physically acceptable
[ 61 , 62 , 64 , 65 ]. Importantly, the stereochemical quality of the structural model is
essential for further studies including molecular simulations.

4 Experimental Constraints to Improve/Verify

Homology Models

Any experimental data that can be used as a structural parameter, even indirectly,
aid in building a better structural model based on homology [ 66 ]. Once a structural
model is available, further experiments can be designed using insights from the
model. Thus, designing experiments using structural models and building models
that satisfy experimental constraints become an iterative process leading to better
understanding of a protein's structure-function relationships. The experimental
constraints that can be used in model building are diverse and we discuss several
examples here. Usually, experimental constraints are sparse and by themselves
not enough to lead to an unambiguous structural model. Thus, several models
can satisfy a given set of constraints. However, the subset of models that do
not satisfy a given experimental constraint can be eliminated from consideration.
The experimental constraints can either be at the residue level or provide overall
structural information. Some of the residue-level constraints include distance
bounds between specific residues obtained by FRET and site-directed cross-linking.
Iterative model building is possible using FRET and site-directed cross-linking,
since a structural model allows probing a much smaller subset of residue-pairs
for distance measurements as opposed to residue-pairs being chosen randomly
[ 2 , 67 ]. Furthermore, these distance measurements provide direct validation of
a given structural model. Residue accessibilities obtained by EPR spectroscopy
[ 17 ] and H-D exchange mass spectrometry [ 18 , 19 ] also aid in model refinement.
Experiments that provide information on the overall protein structure include
small angle X-ray scattering (SAXS) [ 68 ], cryo-electron microscopy (CryoEM)
[ 69 ], and circular dichroism (CD) spectroscopy, among others. SAXS provides
the molecular envelope or the overall shape of the protein in solution, which can