2012 Volume 6 Issue 5 Pages 234-240
A coumarin-based prodrug system plays a significant role in preparing esterase-sensitive prodrugs of amines and peptides. The electronic structures of 27 coumarin-based prodrugs developed in our lab were calculated at a B3LYP/6-31+G (d,p) level with a Gaussian 03 establish five novel QSPR models. The SMLR linear model (q2 = 0.427, r2 = 0.516) and the PLS linear model (q2 = 0.584, r2 = 0.663) were developed with descriptors selected by an Unsupervised Forward Selection method. Another three nonlinear QSPR models were established by a Polynomial Neural Network (PNN) Simulation method (q2 = 0.692, 0.675, 0.663; r2 = 0.700, 0.688, 0.672). We suggest that the QSPR models derived here, especially the PNN models, can be used to predict the release kinetics of coumarin-based prodrugs as well as design new derivatives of coumarin-based prodrug candidates.