In the present work multiple linear regression analyses were performed to build QSAR models for nucleoside analogous using density functional theory (DFT) and molecular mechanics (MM+) based descriptors in both gas and solvent phases. The QSAR models for 14 carbocyclic analogues of nucleosides against murine leukemia cell line (L1210/0) and human T-lymphocyte cell lines (Molt4/C8 and CEM/0) explain more than 90% of the variances in the activity data along with higher values of \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ r_{CV}^2\left( { > 0.86} \right) $$\end{document}. The energy of the next lowest unoccupied molecular orbital (ENL), electrophilicity (ω) and van der Waals surface area (SA) are the main independent factors contributing to the anticancer activity of nucleoside analogues. Inclusion of solvent medium increases the correlation of each descriptor with activity. Based on the key features responsible for anticancer activity, 10 new compounds with rather high anticancer activity have been theoretically designed. Cytotoxic activities of an additional set of 20 nucleoside analogues were also modeled by the same descriptors and found their predicted values to be in good agreement with the experimental values.
