基于膜雌激素受体(GPER)结合化合物能力的分类预测模型  

Classification prediction model based on GPER binding ability of membrane estrogen receptor

作　　者：王宇飞 曹慧明 梁勇[1] WANG Yufei;CAO Huiming;LIANG Yong(Hubei Key Laboratory of Environmental and Health Effects of Persistent Toxic Substances,Institute of Environment and Health,Jianghan University,Wuhan,430056,China)

机构地区：[1]持久性有毒污染物环境与健康危害湖北省重点实验室,环境与健康研究院,江汉大学,武汉430056

出　　处：《环境化学》2022年第2期417-428,共12页Environmental Chemistry

基　　金：国家自然科学基金(21806058)资助。

摘　　要：近年来,计算毒理学的方法被广泛应用于潜在的环境内分泌干扰物(EDCs)的筛选.膜雌激素受体(GPER),作为一种可以快速响应内源性配体雌激素的关键靶蛋白,调控其介导的多项生理学功能.但是针对GPER的化合物毒性预测模型仍未见报道.因此,本研究收集了130个化合物对GPER的结合活性数据,主要包括双酚类、多溴联苯类以及农药杀虫剂类环境污染物.利用随机森林(RF)、支持向量机(SVM)、人工神经网络(ANN)、K最近邻(KNN)、朴素贝叶斯(NB)以及逻辑回归(LG)等6种机器学习算法构建二分类模型.结果显示,所有被测试算法的测试集准确率均达到85%以上,其中SVM、RF、ANN、KNN等4种算法的训练集准确率高于93%,10折交叉验证准确率高于80%,说明得到的模型具有优秀的分类预测性能.因此,本研究基于机器学习算法构建的分类模型,可以用来快速、准确地预测环境污染物是否通过结合GPER产生内分泌干扰效应.为评估环境污染物的潜在健康风险提供了理论依据.In recent years,computational toxicology has been widely applied to the screening of potential environmental endocrine disruptors(EDCs).Membrane estrogen receptor(GPER),as a key target protein that can rapidly respond to endogenous ligand estrogen,regulates its mediated multiple physiological functions.However,the prediction model of compound toxicity for GPER has not been reported.Therefore,the binding activity data of 130 compounds against GPER were collected in this study,mainly including bisphenols,polybrominated biphenyls and environmental pollutants like pesticides and insecticides.Six machine learning algorithms,including random forest(RF);support vector machine(SVM),artificial neural network(ANN),K-nearest neighbor(KNN),Naive Bayes(NB)and logistic regression(LG),were used to construct the dichotomous model.The results showed that the test set accuracy of all the algorithms tested reached more than 85%,among which the training set accuracy of SVM、RF、ANN and KNN algorithm was 93%higher than that of the four algorithms,and the 10-fold cross validation accuracy was 80%higher than that of the four algorithms,indicating that the model obtained had excellent classification prediction performance.Therefore,the classification model built based on machine learning algorithm in this study can be used to quickly and accurately predict whether environmental pollutants will produce endocrine disrupting effects by combining with GPER.It provides a theoretical basis for evaluating the potential health risks of environmental pollutants.

关 键 词：内分泌干扰物 膜雌激素受体 机器学习算法 分类预测模型 

分 类 号：R114[医药卫生—卫生毒理学] X503.1[医药卫生—公共卫生与预防医学]