基于药理学网络模型的抗肿瘤药物不良事件预测

Predicting Cancer Drug Adverse Events Based on Pharmacological Network Model

目的:针对抗肿瘤药物引起的不良事件,为提高患者的生活质量,提出了一种抗肿瘤药物不良事件的预测方法,从而减少药品不良事件的发生。方法:该方法选择了药理学网络模型(pharmacological network models,PNM),在充分考虑时间顺序的基础之上,由特定药物和不良事件信息的关联构建二分网络,定义3类协变量,采用逻辑回归实现预测。文中选择美国食品药品监督管理局不良事件报告系统(FAERS)数据库2010年的数据,构建了由151种抗肿瘤药物和625种不良事件组成的网络,通过训练逻辑回归模型对2011～2015年FAERS数据库中的新抗肿瘤药物-不良事件关联组合进行预测。结果:PNM实现了受试者工作特征曲线下面积(AUROC)为0. 824,具有良好的预测结果。结论:PNM对抗肿瘤药物的不良事件有良好的预测性能,可以为临床的合理用药以实际指导意义。

Objective:In order to improve the living quality of tumor patients,a method for predicting adverse events of anticancer drugs was proposed,and this method can reduce the occurrence of adverse drug events.Methods:In this paper selecting the Pharmacological Network Models(PNM),which taked into account the chronological order,a bipartite network was constructed by the known specific drugs and adverse drug events associations,and then prediction was implemented by logistic regression based on the definition of three types of covariates.According to the FAERS 2010 database,we constructed a network representation of drug-ADE associations for 151 drugs and 625 ADEs,a logical regression model was trained to predict unknown drug-ADE associations in 2011-2015 FAERS database that were not listed in the 2010 FAERS database.Results:PNM achieved an AUROC(area under the receiver operating characteristic curve)of 0.824,which had a good predictive performance for drug adverse events.Conclusion:This method can be used as a practical guiding significance for clinical rational drug.