| |
Abstract:
A series of reported recently oxazole-4-carboxamides as
the inhibitor of glycogen synthase kinase-3 (GSK-3) were
firstly used to establish quantitative structural
activity relationship models. Heuristic method was
applied to select efficient molecular descriptors. At
the same time, a linear model with correlation
coefficient (R=0.93) and cross validation correlation
coefficient (RCV=0.90) was built. In addition, gene
expression programing and back-propagation artificial
neural network methods were carried out to establish
nonlinear models. As a result, a five molecular
descriptors nonlinear model with correlation coefficient
(R) and mean square error (MSE) of 0.95 and 0.15 for the
training set, 0.95 and 0.17 for the test set was
developed by GEP. Moreover, a more robust nonlinear
model with of 0.97 for the training set, 0.95 for the
test set, 0.89 for the validation set and 0.96 for the
whole data set was built by BPANN. As we can see, both
GEP and BPANN are promising methods to explore nonlinear
relationships for good statistical performance.
Expectantly, these models can give assistance to
discover and design new inhibitors of GSK-3 in the
future.
|
|
|
