Biomedical Engineering Reference
In-Depth Information
Three classes of descriptors: electrotopological, thermodynamic, and structural
were considered for statistical fitting of NCC blockers using the GFA methodology,
implemented in Cerius2 software [
73
]. The electrotopological descriptors are
numerical values computed for each atom in a compound and which encode
information about both topological environments of that atom and electronic
interactions due to all other atoms in the compound. The topological relationship
is based on the graph distance to each atom. The electronic aspect is based on
intrinsic state and perturbation due to intrinsic state differences between atoms in
the compound [
74
,
75
]. Thermodynamic descriptor describes energy of compounds
and their conversions. Selection of descriptors is based on their correlation with
NCC blocking activity and capability of producing multiple linear regression
models with moderate correlation coefficients. A statistically meaningful QSAR
model was constructed by checking the variation of different statistical parameters
against the number of descriptors. Details of the descriptors, QSAR equations, and
analysis are available in our previously published work [
71
].
The QSARmodel was created on a training set of 83 NCC blockers, and validated
by a test set of 21 NCC blockers. The model was developed using Atype_C_24,
Atype_N_68, Rotlbonds, S_sssN, and ADME_Solubility information-rich
descriptors, which play an important role in determining NCC blocking activity.
Among these, Atype_C_24, Rotlbonds, and S_sssN descriptors correlated positively
with activity, while Atype_N_68 and ADME_Solubility descriptors correlated
negatively with NCC blocking activity. These descriptors provided insight into the
physicochemical requirements necessary for designing potent NCC blockers.
The robustness of the developed 2D-QSAR models obtained by GFA method was
evaluated using a cross-validation strategy involving different techniques. This model
was further validated using the leave-one-out cross-validation approach, Fischer
statistics, Y-randomization test, and predictions based on the test data set [
76
,
77
].
Y-randomization test confirms whether the model is obtained by chance correlation,
and is a true structure-activity relationship to validate the adequacy of the training set
compounds. The resulting descriptors produced by QSARmodel were used to identify
physicochemical features relevant to NCC blocking activity. The predictive properties
of the developed model were more rigorously tested by predicting the NCC blocking
potency of external test set of 21 compounds, details available elsewhere [
71
].
2D-QSAR model is statistically significant and explains more than 95% of
the variance in the actual (experimental) activity with good predictive power [
71
].
The structural and electrotopological index descriptors were found to play a major
role in determining NCC blocking activity. The atom-type descriptors and rotational
bond highlight the significance of spatial aspects in designing these blockers. The
influence of electrotopological descriptor S_sssN was promising, and showed that
the tertiary nitrogen with its linear aliphatic amine contributions must be taken into
account when designing new inhibitors against this channel [
71
]. Analysis of atom-
wise contributions to hydrophobicity will probably help to appropriately take into
account those atom types that are essential for determining NCC blocking activity.
Type, definition, and meaning of these five descriptors are described below
along with their
importance in understanding the NCC blocking activity.
