人工神经网络法预测核苷及核酸碱基的疏水分配系数

Prediction of Hydrophobic Partition Coefficients of Nucleosides and Nucleoside Bases by Artificial Neural Network

用人工神经网络方法预测核苷及核酸碱基一类化合物的ｌｇＰ值（Ｐ：１－辛醇／水分配系数），预测精度显著优于ＢｌｇＰ法、ＣｌｇＰ法和ＡｌｇＰ法。根据预测结果讨论了分子内氢键及分子构象柔顺性对这类化合物疏水性的影响。

A back-propagation artificial neural network (ANN) was trained on the Gramer's parameters to predict the lgP (logarithm of '1-octanol to water' partition coefficients) of nucleosides and nucleoside bases. The prediction results show that the ANN method is betterthan other methods (BlgP, ClgP, AlgP) for this class of compounds.The effect of conformational flexibility or intramolecular hydrogen bonding on the lgP of nucleoside compounds is discussed.