基于启发式变量筛选方法研究与雌激素β受体结合的化合物结构与活性之间的关系

Quantitative structure-activity relationship of compounds binding to estrogen receptor β based on heuristic method

雌激素类化合物由于其对人和野生动物健康的负面影响而受到广泛关注.雌激素受体存在两种亚型(ERα和ERβ),化合物与两种受体亚型在结合活性和化合物结构特征方面存在差异.以31种与雌激素β受体亚型(ERβ)结合的化合物为研究对象,采用启发式变量筛选方法,从1524个变量中筛选出5个与化合物活性(lgRBA)最相关的变量,然后采用多元线性回归(MLR)建立最佳预测模型.模型相关性显著,而且具有良好的稳健性和预测能力(r2=0.829,q2LOO=0.742,r2pred=0.772,q2ext=0.724,RMSEE=0.395).同时揭示了影响化合物与ERβ受体结合的配体化合物分子的结构特征,并对模型的应用域进行了研究.

Estrogen compounds may pose a serious threat to the health of humans and wildlife. The estrogen receptor （ER） exits as two subtypes, ERα and ERβ. Compounds might have different relative affinities and binding modes for ERα and ERβ. In this study, heuristic method was performed on 31 compounds binding to ERβ to select 5 variances most related to the activity （LogRBA） from 1524 variances, which were then employed to develop the best model with the significant correlation and the best predictive power （r2 = 0.829, q2LOO = 0.742, r2pred = 0.772, q2ext = 0.724, RMSEE = 0.395） using multiple linear regression （MLR）. The model derived identified critical structural features related to the activity of binding to ERβ. The applicability domain （AD） of the model was assessed by Williams plot.