基于动态代谢组学数据通路信息校正的混合效应模型建立及模拟研究  

The Establishment of Mixed-effect Model based on Pathway Information to Analyze Dynamic Metabolomics Data and its Simulation Study

作　　者：田伟 张秋菊[1] 侯小文 李称 丛雨欣 赵敏[1] 刘美娜 Tian Wei;Zhang Qiuju;Hou Xiaowen(School of Public Health,Harbin Medical University,150086,Harbin)

机构地区：[1]哈尔滨医科大学卫生统计学教研室,150081

出　　处：《中国卫生统计》2020年第5期678-681,686,共5页Chinese Journal of Health Statistics

基　　金：国家重大专项课题(2016ZX08011005);国家自然基金青年基金(81502889)。

摘　　要：目的建立基于代谢通路信息校正的混合效应模型,通过模拟实验评价模型筛选动态代谢标志物的效果,为动态代谢组学生物标志物筛选提供方法。方法通过条件自回归模型,利用代谢物在通路中的位置关系,结合一阶自回归模型,建立基于代谢通路信息校正的混合效应模型并进行模拟研究;利用模型实现代谢标志物相关关系和时间因素的拟合,结合处理因素回归系数的95%可信区间获得动态代谢标志物。结果建立了基于代谢通路信息校正的混合效应模型;在各种模拟条件下,模型对动态代谢标志物的筛选均有较高灵敏度,特异度以及准确率;动态时间点数和代谢物自身噪声水平对模型的灵敏度有较大影响;通过模拟实验发现时间点数大于5时模型已有较高的拟合精度;样本量是影响模型拟合效果的重要因素,当样本量大于30时,模型筛选差异标志物的灵敏度已大于90%。结论基于通路信息校正的混合效应模型,能够很好地应对动态代谢组学数据分析的各种复杂情况,是动态代谢组生物标志物筛选研究的可选方法之一。Objective Our aim is to establish a mixed effect model adjusted by metabolic pathway information and evaluate its ability for selecting dynamic metabolic biomarker to provide a biomarker screening method for dynamic metabolomics research.Methods The model was established based on the conditional autoregressive model and the positional relationship of metabolites in metabolic pathways,and combined with first-order autoregressive model.Through fitting relationship among metabolites and the influence of time factor,this model could be used to screen biomarker according to their 95%credibility intervals.Results The pathway adjusted mixed effect model was established.Under a variety of simulation conditions,high levels of sensitivity,specificity and accuracy were presented.The numbers of time points and noise levels of metabolites would influence fitting effects,and it would get higher fitting accuracy when numbers of time points is greater than 5.The sample size would affect model′s effect where if it is more than 30,sensitivity of it would be over 90%.Conclusion The pathway adjusted mixed effect model could be able to deal with complicated problems in dynamic metabolomics analysis and is a feasible method for biomarker screening.

关 键 词：动态代谢组学 混合效应模型 代谢通路 标志物 

分 类 号：C81[社会学—统计学]