机器学习预测化合物的雌雄激素受体活性  

Machine learning predicts androgynous receptor activity in molecule compounds

作　　者：胡帅 孔韧 谢良旭 Hu Shuai;Kong Ren;Xie Liangxu(Institute of Bioinformatics and Medical Engineering(School of Electrical and Information Engineering,Jiangsu University of Technology),Changzhou 213001)

机构地区：[1]江苏理工学院电气信息工程学院生物信息与医药研究工程研究所,常州213001

出　　处：《现代计算机》2023年第1期16-21,共6页Modern Computer

基　　金：国家自然科学基金面上项目(22003020,81603152,81903661)。

摘　　要：预测一种化合物的雌雄激素受体的活性,对于避免暴露于环境中类似雌雄激素的化学物质是非常重要的。采用包括支持向量机、随机森林等多种机器学习方法,利用Binding Database数据库建立了预测活性的定量结构-活性关系(quantitative structure-activity relationship,QSAR)模型,并对模型进行验证与评估。评估结果发现,随机森林结合扩展连通性分子指纹对数据集预测准确率为0.83,其受试者工作特征曲线(receiver operating characteristic curve,ROC)得到的数据集曲线下面积(area under curve,AUC)为0.892,表明该模型具有广泛且良好的预测能力。该研究建立的活性预测模型可用于化合物的活性预测,为内分泌化合物的活性评估和风险管理提供参考。Predicting estrogen and androgen receptor activity of a small molecule compound is important to avoid exposure to environmental androgynous chemicals.In this paper,a Quantitative Structure-Activity Relationship(QSAR)model for predicting Activity was established by using Binding Database,including support vector machine,random forest and other machine learning methods.The model is verified and evaluated.The evaluation results show that the prediction accuracy of random forest combined with extended connectivity molecular fingerprint on the dataset is 0.83.The Area Under curve(AUC)of the dataset obtained from the Receiver Operating Characteristic curve(ROC)was 0.892,indicating that the model had good and extensive predictive ability.The activity prediction model of androgynous receptor established in this study can be used to predict the activity of compounds,in order to provide a reference for the activity assessment and risk management of endocrine compounds.

关 键 词：雌雄激素受体 QSAR 活性预测 机器学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]