QSAR Modeling of HIV‐1 Protease Inhibition on Six‐ and Seven‐membered Cyclic Ureas

Abstract

A Quantitative Structure -Activity Relationship (QSAR) analysis was carried out on a dataset of 135 six-and seven-membered cyclic urea-based Human Immunodeficiency Virus Type 1 (HIV-1) protease inhibitors. Using a larger and more diverse dataset than previous studies reported in literature allowed a more comprehensive analysis. A large set of molecular descriptors, calculated with CODESSA PRO, was used in Multiple Linear Regression (MLR) analysis and the resulting four-parameter model enabled accurate prediction of inhibitory activity for the structures both in training and external validation sets. The squared correlation coefficient for both sets combined is 0.824. Descriptors in the final model characterize the size and shape of the molecule and its charge distribution. Non-linear relationships between the HIV-1 protease inhibition activity and some descriptors are significant and described with simple non-linear transformations. After the model development, another dataset of 45 compounds was collected and additional external validation was performed.

The QSAR Modeling of Cytotoxicity on Anthraquinones

“…This left 349 descriptors for the analysis, which were then subjected to nonlinear transformations. Transformations were necessary in order to still use the linear regression and at the same time also account for the possible nonlinear effects between cytotoxicity and structure [20]. This increased the number of descriptors to 1538 including 349 original untransformed descriptors: 227 logarithmic descriptors (indicated in descriptor names as LOG); 264 reciprocal ones (1/); 349 squared (SQ) and 349 square-root versions (SQRT).…”