Most of the current methods search for a binding site using a
static protein structure and the calculated values of the various
geometrical descriptors contribute to the final value of the drugg-
ability index. TypicalIDvalues are in the range of3.0 to 0.0
[8]. Binding pockets with anIDgreater than1.0 are having a
high potential for being druggable, whereas pockets with anIDless
than1.5 are having a low potential for small-molecule drugg-
ability [8]. The druggability index can be calculated using the
Eq. (1):ID¼XNi¼ 1½ðaiXiþbilogðÞXi 1 ÞWhereNis the number of the active-site descriptors used,Xiis
theith descriptor, andaiandbiare coefficients (obtained from
fitting to experimental data) for the linear and logarithmic terms
of theith parameter, respectively [8].
Cheng et al. defined the terms, “curvature” and “lipophilicity,”
which reflect the shape of the binding site and its hydrophobicity,
respectively. These two parameters indicate the maximal binding
affinity for drug-like compounds [35–37]. Other binding sites’
identification methods use different geometrical descriptors. An
example is the Dscore method, which forms a linear function of
three pocket properties: size, enclosure (akin to an average degree
of buriedness), and hydrophilicity [11, 38]. Another druggability
evaluation method, which originated from NMR-Based fragment
screening, try to mimic the hit rates concept, This hit rate method
relies on the assumption that sites that bind with a higher propor-
tion of fragments are also more likely to deliver high-affinity, non-
covalent drug-like leads [3]. The FTMAP is an example of such
methods and it can estimate the maximal binding affinity, while
evaluating the druggability of a given target toward a particular
compound.2.3 Importance
of Protein Flexibility
All the methods described above do not take the protein flexibility
into account, making them very limited in identifying and studying
complicated and hidden binding sites (seeNote 2). For example,
many studies have shown that the druggability index can vary
significantly, when calculating the same index for the same target,
but for a different state (i.e., free vs. bound structures) [8]. This is
clearly demonstrated in highly dynamical proteins [39]. For exam-
ple, the unbound Bcl-xL protein cannot accommodate the Bak
peptide, which is normally associated with a low druggability
index (ID¼2.0). On the contrary, the bound conformation is
much wider and deeper, which is reflected by a much higher drugg-
ability index (ID¼0.5) (Fig.4)[8].92 Tianhua Feng and Khaled Barakat