构建有机化合物斑马鱼雌激素干扰效应的二元分类模型

计算毒理学方法已成为辅助内分泌干扰物(EDCs)管理的决策支持工具。因此,发展内分泌干扰效应指标的(定量)结构活性关系((Q) SAR)等预测模型对于实现EDCs环境管理具有重要的意义。在雌激素受体(Q) SAR模型研究方面,目前主要针对人、牛、大鼠和小鼠等物种的雌激素受体干扰效应进行了研究,而对鱼等水生生物雌激素受体干扰效应等指标的(Q)SAR模型研究还较少。本研究采用基于欧几里德距离的K最近邻(k NN)分类算法,构建了斑马鱼雌激素受体干扰效应的二元分类模型。结果表明,2个最优模型训练集和验证集的预测准确度(Q)、敏感性(Sn)和特异性(Sp)参数均大于0.93,说明模型具有较好的预测能力。因此,能够用所建模型填补模型应用域内其他化合物缺失的斑马鱼雌激素受体干扰效应定性数据。

Development of Binary Classification Models for Predicting Estrogenic Activity of Organic Compounds on Zebrafish

Computational toxicology method has been a critical decision support tool for the management of endocrine disrupting chemicals (EDCs). Thus, it is of vital importance to develop the predictive models e.g. (quantitative) structure-activity relationship ((Q)SAR) for EDCs environmental management. For the available (Q)SAR models for estrogen receptor up to now, only the estrogenic activity of four species such as calf, rat, mouse, and human were modeled. While the (Q)SAR models for aquatic organisms e.g. fish were still less. Here, the binary classification models for predicting the estrogenic activity of zebrafish was attempted to construct employing the K nearest neighbor (kNN) classification algorithm based on Euclidean distance. The modeling results indicated that all of the prediction accuracy (Q), sensitivity (S_n) and specificity (S_p) of training sets and validation sets for the two optimal models are greater than 0.93, indicating that the models have good prediction ability. Therefore, the missing estrogenic activity data gap for other organic compounds within the application domain of derived models on their missing estrogenic activity data could be filled.