Comparative Analysis of Techniques Oriented on the Recognition of Ligand Binding Area in Proteins - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 82

to a specific group of proteins) indicates that binding pockets emerge through many
different mechanisms. This phenomenon could be related to the diverse biological
activity attributed to proteins with FMN and NAD + cofactors.
The structural similarity of binding pockets - as indicated by geometry-bases
tools - is clearly discernible, although limited in scope.
In the FOD method, the most important factor affecting the identification of
binding pockets (by searching for residues which represent local distortions in the
structure of the protein's hydrophobic core) appears to be the presence of additional
polypeptide chains determining the protein's quarternary structure. Of nearly equal
importance is the size of polypeptide chains: large single-core proteins can be sub-
divided into domains where the hydrophobic core structure exhibits better ordering
than in the case of a multi-domain multi-core molecule. The phenomenon of ligands
being bound in the inter-chain space (between adjacent subunits) also affects the
accuracy of predictions returned by the FOD model.
Given the presented results, it seems valid to conclude that binding pockets are
generated through many different mechanisms and that the applicability of each
theoretical model is typically limited to a specific subset of proteins, yielding poor
results for proteins which do not share the preferred structural properties - even if
their general structure (or purpose) is similar.
The set of programs discussed in this chapter was also tested in the context of
active site identification in hydrolases (Prymula et al. 2011 ) . In that study, knowledge-
based tools proved more reliable than geometry-based packages. In contrast, geometry-
based software appears to return better results for FMN and NAD + binding site
identification, as described above. We can therefore conclude that the binding geom-
etry of these ligands is highly deterministic and specific, whereas evolutionary
factors play a more pronounced role in shaping enzymatic active sites. Distortions
of the protein's hydrophobic core are more closely related to ligand binding sites
(Fig. 4.18 ). It appears that the structure of FMN and NAD + pockets is local in character
and does not affect the shape of molecule as a whole - contrary to active centers in
hydrolases.
The comparison presented in Fig. 4.18 explains some of the difficulties involved
in identifying ligand binding residues by way of the FOD model. The Δ H pro fi le indi-
cates that the hydrophobicity attributed to such residues remains in agreement with
statistical predictions and therefore does not trigger distortions in the protein's core.
This is especially evident in the case of 3F2V (Fig. 4.18 ), where the placement of
the active site (and thus of the substrate) is tied to a specific deformation and can
therefore be accurately identified.
The presented examples also point to the role of the environment surrounding the
ligand binding and catalytic residues. Enzymatic active sites need to be shielded
from water and are usually located in deep pockets, where their hydrophobicity
deficiency can be clearly discerned on the

Δ H plot (see protein 2Z6I in Fig. 4.17 ).
Binding ligands involves compensating the protein's own hydropbhobicity
deficiencies with the ligand's own excess hydrophobicity, resulting in a droplike
core structure. In some cases, however (as depicted in Fig. 4.18 , protein 1BMD), the
ligand may bind at locations with variable hydrophobicity conditions. Since certain