Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT

ABSTRACT - Internet Electron. J. Mol. Des. October 2004, Volume 3, Number 10, 622-650

Quantitative Structure-Activity Relationship Study of Bisphosphonates Aihua Xie, Chenzhong Liao, Zhibin Li, Zhiqiang Ning, Weiming Hu, Xianping Lu, Leming Shi, and Jiaju Zhou Internet Electron. J. Mol. Des. 2004, 3, 622-650

Free: Download the paper in PDF format

Return to Table of Contents

Get Acrobat Reader to view and print the paper

Abstract:
Bisphosphonates (BPs) are most widely used as agents for treating osteoporosis. They have also been used for other purposes such as herbicides, anticancer agents, and antiparasitics. Here we report QSAR models of four BPs datasets based on the 118 structural and biological data we have collected from various literature sources. The structures were energy minimized under MMFF94 force field, then 181 2D and 3D descriptors were calculated with MOE (Molecular Operating Environment). The step by step multiple regression and principle analysis were used to achieve our QSAR models. The predictive power of each QSAR model was estimated with the leave-one-out (LOO) cross-validation method. The QSAR model for a dataset of 28 GGPPSase inhibitors (r² = 0.86, r²_LOO = 0.82, s = 0.45, s_LOO = 0.51, F = 77.56) is made up of two descriptors. Another dataset of 28 compounds with bioactivities against the growth of T. Brucei rhodesiense was studied using PCA and reached a model (r² = 0.85, r²_LOO = 0.79, s = 0.30, s_LOO = 0.35, F = 32.03) with four principle components (PCs). Both the above models have comparable predictive ability with CoMFA model reported by Szabo et al. The 86 BPs provided by Novartis with in vivo bio-data of TPTX rats were divided into two datasets. A six PCs model (r² = 0.80, r²_LOO = 0.72, s =0.44, s_LOO = 0.53, F = 24.83) elucidated the dataset of 44 compounds in which containing aliphatic linked nitrogen atoms. The other dataset includes 42 BPs containing a heterocyclic moiety with at least one nitrogen atom. Its PCA model (r² = 0.80, r²_LOO = 0.71, s = 0.46, s_LOO = 0.57, F = 19.99) consists of seven PCs. A leave-four-out test procedure shows that though the QSAR models based on in vivo bone resorption pED₅₀ values cannot provide explicit indications for drug design, their predictive ability for related compounds is quite good.

Free: Download the paper in PDF format

Return to Table of Contents

Get Acrobat Reader to view and print the paper