摘要
目的系统评估因果成分中介模型(causal compositional mediation model,CCMM)、等距对数比变换的微生物组中介分析(isometric log-ratio transformation for microbiome mediation,Isometric LRTMM)和稀疏微生物组因果中介模型(sparse microbial causal mediation model,Sparse MCMM)3种微生物组成分中介分析方法的性能,探索肠道菌属在克罗恩病不同治疗方式与治疗反应中的潜在中介作用。方法通过模拟研究比较3种方法在不同样本量和中介变量维度下的效应估计精度和变量筛选能力,基于加拿大3所医院的儿童克罗恩病患者数据,分析治疗方式、肠道菌属与治疗反应间的关联。结果模拟结果显示:在效应估计方面,各情形下CCMM均优于Sparse MCMM,其总效应、直接效应及间接效应偏倚、方差、均方根误差均最小;中介变量筛选方面:Isometric LRTMM表现最佳,CCMM次之,Sparse MCMM较弱。应用研究识别出5种与治疗反应相关菌属(粪球菌属、肠杆菌属、别样杆菌属、萨特氏菌属、柯林斯菌属),但均未发挥中介作用。结论微生物组成分中介分析时,若侧重效应值估计推荐CCMM,若关注中介识别则Isometric LRTMM更优;本研究发现5种克罗恩病治疗反应相关菌属,为益生菌治疗靶点的筛选提供依据。
Objective To evaluate the performance of three mediation analysis methods for microbial compositional data[causal compositional mediation model(CCMM),isometric log-ratio transformation for microbiome mediation(IsometricLRTMM),and sparse microbial causal mediation model(SparseMCMM)],and to explore the potential mediating roles of microbial genera in the associations among treatment modalities for Crohn's disease and treatment response.Methods Simulation studies were conducted to quantify the bias,mean squared error,and variance of direct and indirect effects estimates for three mediation analysis methods under varying sample sizes and mediator dimensions scenarios.Then further applied three methods to real-world data to investigate the relationships among treatment strategies,the gut microbiome,and treatment responses in Crohn's disease.Results The simulation study showed that in mediation effect estimation,CCMM consistently outperformed SparseMCMM across all scenarios,yielding the smallest bias,variance,and root mean squared error for total,direct,and indirect effect.For mediator selection:IsometricLRTMM exhibited the best performance,followed by CCMM,while SparseMCMM performed less favorably.Application to real data identified five genera associated with treatment response(Faecalibacterium,Enterobacter,Parabacteroides,Sutterella,and Collinsella),none of which demonstrated a mediating role.Conclusions CCMM is recommended when accurate effect estimation is the priority,whereas IsometricLRTMM is preferable for identifying microbial mediators.This study highlights five genera associated with treatment response in Crohn's disease,providing potential targets for the probiotic-based therapeutic strategies.
作者
王菊平
和思敏
李瑞凡
司鑫雨
曹红艳
王彤
WANG Juping;HE Simin;LI Ruifan;SI Xinyu;CAO Hongyan;WANG Tong(Department of Health Statistics,School of Public Health,Shanxi Medical University,Taiyuan 030001,China;MOE Key Laboratory of Coal Environmental Pathogenicity and Prevention,Shanxi Medical University,Taiyuan 030001,China;Department of Mathematics,School of Management,Shanxi Medical University,Taiyuan 030001,China;Department of Medical Information Retrieval,School of Management,Shanxi Medical University,Taiyuan O30001,China)
出处
《中华疾病控制杂志》
北大核心
2026年第3期324-331,共8页
Chinese Journal of Disease Control & Prevention
基金
国家自然科学基金(82073674,82373692)
山西省基础研究计划资助项目(202103021223234)。
关键词
成分数据
中介分析
微生物组数据
Compositional data
Mediation analysis
Microbiome data
