作者: Siavash Riahi , Eslam Pourbasheer , Rassoul Dinarvand , Mohammad Reza Ganjali , Parviz Norouzi
DOI: 10.1111/J.1747-0285.2009.00843.X
关键词:
摘要: The quantitative structure-activity relationship of the novel 6-naphthylthio 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio) thymine derivatives for prediction anti-human immunodeficiency virus type 1 activity was studied. suitable set molecular descriptors calculated and important using variable selections stepwise multiple linear regression genetic algorithm were selected. A comparison between attained results indicated superiority over method in feature-selection. predictive quality models tested an external eight compounds, randomly chosen out 39 compounds. algorithm-multiple model with six selected obtained. This model, demonstrating high statistical qualities (R(2)(train) = 0.925, Q(2) 0.872, SE (%) 1.23, F 49.338, R(2)(pred) 0.944), could predict molecules a error percentage lower than 10%. suggest that electronegativity, masses, atomic van der Waals volumes are main independent factors contributing to studied