Development of Predictive Quantitative Structure-Activity Relationship Models of Epipodophyllotoxin Derivatives

Abstract

Epipodophyllotoxins are the most important anticancer drugs used in chemotherapy for various types of cancers. To further, improve their clinical efficacy a large number of epipodophyllotoxin derivatives have been synthesized and tested over the years. In this study, a quantitative structure-activity relationship (QSAR) model has been developed between percentage of cellular protein-DNA complex formation and structural properties by considering a data set of 130 epipodophyllotoxin analogues. A systematic stepwise searching approach of zero tests, missing value test, simple correlation test, multicollinearity test, and genetic algorithm method of variable selection was used to generate the model. A statistically significant model (r((train))(2) = 0.721; q(cv)(2) = 0.678) was obtained with descriptors such as solvent-accessible surface area, heat of formation, Balaban index, number of atom classes, and sum of E-state index of atoms. The robustness of the QSAR models was characterized by the values of the internal leave-one-out cross-validated regression coefficient (q(cv)(2)) for the training set and r((test))(2) for the test set. The root mean square error between the experimental and predicted percentage of cellular protein-DNA complex formation (PCPDCF) was 0.194 and r((test))(2) = 0.689, revealing good predictability of the QSAR model.