期刊名称：Beni-Suef University Journal of Basic and Applied Sciences
印刷版ISSN：2314-8535
电子版ISSN：2314-8543
出版年度：2016
卷号：5
期号：4
页码：320-333
DOI：10.1016/j.bjbas.2016.11.003
语种：English
出版社：Elsevier
摘要：Abstract The research entails the development of two models that can be used to predict the toxicity of compounds suspected of having anti-leukemia activity, the two models, one for MOLT-4 and the other for P388 cell lines, were developed by carefully selecting 5 descriptors in each case from a pool of 4900 molecular descriptors using GA-MLR method. The ability of the models to predict chemicals that were not used in the preparation set was evaluated by validating the models internally [Q2 CV MOLT-4 = 0.644 and Q2 CV (P388) = 0.641] and externally [R2 pred (P388) = 0.662 and R2 pred (MOLT-4) = 0.679], other criteria such as r2 m, R2 p, k and k′ were also determined and were found to support R2 pred results of the models. The effects of every descriptor in the model were dictated by calculating their individual mean effect to (pLC50) the toxicities of the compounds. SpMax7_Bhm (1.697) and RDF145u (−4.300) influenced the models the most, while in MOLT-4 the descriptors with the least effect was n-isopropyl (0.0343) and for P388 is 1.162. The models were further validated by subjecting them to y-randomization which finally confirmed the robustness of the models developed.
