Chemistry Reference
In-Depth Information
a
b
c
φ
d
van der waals
H-bond
Electrostatic
Distance (
r
ij
)
Distance (
r
ij
)
Distance (
r
ij
)
Fig. 4 (a) A representation of p38 mitogen-activated protein kinase structure bound to BIRB796
and (b) an expanded view of the binding site. (c) A representation of the hydrogen-bonding (
red
)
and electrostatic interactions (
green
) between the atoms of the protein and the atoms of the ligand.
(d) A representation of three force-field energy terms (van der Waals, hydrogen-bonding, and
electrostatic) as distance between the interacting atom pairs change. (Reprinted with permission
from [
30
], copyright 2004 by the Nature Publishing Group)
to reproduce known experimental structures, and the binding score generated has
little relevance to an actual binding affinity. This is an issue similar to empirical
scoring functions; the accuracy of the scoring function is strongly dependent on the
similarity of the protein-ligand complex to the training data set.
As implied, shape-based scoring functions are based on a shape match between
the ligand and the ligand binding site [
110
]. These scoring functions are typically
used as prefilters to eliminate compounds that are unable to fit into the ligand
binding site [
111
,
112
]. Shape-based scoring functions are very fast, but are limited
relative to more accurate scoring functions that calculate binding affinities. Shape-
based scoring functions typically generate smooth energy surfaces using Gaussian
functions [
111
], which are more tolerant to atomic variations and make protein
