SVM Classification and CoMSIA Modeling of UGT1A6 Interacting Molecules

The human UDP-glucuronosyltransferase
1A6 (UGT1A6) plays important
roles in elimination of many xenobiotics, including drugs. We have
experimentally assessed inhibitory properties of 46 compounds toward
UGT1A6 catalyzing the glucuronidation of 1-naphthol and built models
for predicting compounds interactions with the enzyme. The tested
compounds were divided into a training set (n = 31;
evaluated by 10-fold cross-validation) and an external test set (n = 15), both of which yielded similar accuracies (80–81%)
and Matthews correlation coefficients (0.61–0.63) when classified
using support vector machines. Comparative molecular similarity index
analysis (CoMSIA) modeling was conducted for nine methods of compound
alignment. The most predictive CoMSIA model was analyzed in the light
of a homology modeled UGT1A6 structure, with leave-one-out cross-validation,
yielding a q2 of 0.62 and r2 of 0.91 on the training set and a r2pred of 0.82 on the test set. The CoMSIA contour
plots highlighted the importance of H-bond donors and electrostatic
field interactions, accounting for 28% and 25% contribution of the
model, respectively.