有机化合物水溶解性的QSPR研究

Study on the Quantitative Structure Property Relationship of Aqueous Solubility of Organic Compounds

将自相关拓扑指数作为分子描述符 ,研究了多氯联苯类化合物的水溶解性与分子结构的定量关系。将遗传算法 (GA)引用于本研究中的特征提取 ,并将神经网络法用于数学建模 ,与相应体系的水溶解性进行相关分析时发现 ,自相关拓扑指数能较好地反映化合物的结构特征 ,基于GA算法提取的参量用于神经网络法研究得到较优结果。

The autocorrelation of topological index being used as molecular descriptor, Quantitative structure\| property relationship between the aqueous solubility of polychlorinated biphenyls and molecular structure is studied. Genetic algorithm(GA) is used to extracted the feature, and the feed\|forward neural networks are used to develop the correlation model. The results demonstrate that the property of compounds could be described better by the topological index. Superior results have been achieved by quasi\|Newton neural network to regression analysis.