摘要
采用量子化学从头算法,计算10个大黄素类似物分子结构参数,用逐步回归分析法筛选合适结构参数作为网络输入节点,建立大黄素类化合物抗乳腺癌(MCF-7)细胞活性的BP神经网络模型。研究表明,大黄素类化合物抗乳腺癌(MCF-7)细胞活性与分子中A环电荷和第3条最低空轨道能量相关,神经网络模型经留一法交叉验证,具有可信的预报能力。
The molecular indexes of 10 compounds were calculated on ab inito method of quantum chemistry, Then the relative indexes were selected by stepwise regression analysis （SRA）. The quantitative structure-activity relationship of emodin derivatives against the cell line of MCF - 7 was studied on the artificial neural network（ANN）. The study shows the activity of emodin derivatives against the cell line of MCF- 7 is related to the negative charge of A circle and energy of the third low-unoccupied molecular orbit. The ANN model built up in this paper has the reliable predictability by validation on leave-one-out method.
出处
《陕西理工学院学报（自然科学版）》
2007年第4期64-67,共4页
Journal of Shananxi University of Technology:Natural Science Edition
基金
陕西理工学院专项科研项目(SLGQD0508)
