肠道微生物测序研究中不同实验来源差异的综合评价以及应用

Comprehensive Evaluation and Application of Experimental Sources of Variation in Gut Microbiome Sequencing Studies

肠道微生物测序研究具有将微生物结果转化为人类健康的巨大潜力。16S扩增子测序和宏基因组鸟枪测序(whole-metagenome shotgun,WMS)是微生物组研究中的两大主要方法,各具优势。然而,研究样品的异质性、测序仪和文库制备方法的差异如何影响肠道微生物测序结果的可重复性仍有待深入研究。该研究旨在通过比较粪便样本中微生物组成的差异,为测序技术的选择提供参考标准。3种广泛采用的测序仪的结果显示,WMS法中技术重复相关性(r=0.94)较高,而生物学重复相关性(r=0.69)较低。Bray-Curtis距离表明,生物学重复的差异大于技术重复(P<0.001)。此外,16S和WMS数据集间具有明显的分类学图谱差异。研究结果表明,同质化是样品DNA提取前的一个必要步骤;测序仪对分类学变异的贡献小于文库制备方法。我们开发了经验贝叶斯方法,即在计算中“借用信息”,利用标准化数据和(非)参数先验分布分析批次效应参数,提高了16S和WMS之间的群体可比性,为进一步应用于融合分析已发表的16S和微生物数据集提供了依据。

Gut microbiome sequencing studies have great potential to translate microbial analysis outcomes into human health research.Sequencing strategies of 16 S amplicon and whole-metagenome shotgun(WMS)are two main methods in microbiome research with respective advantages.However,how sample heterogeneity,sequencers and library preparation protocols affect the sequencing reproducibility of gut microbiome needs further investigation.This study aims to provide a reference for the selection of sequencing technologies by comparing differences in microbial composition from different sampling sites.The results of three widely adopted sequencers showed that the technical repetition correlation(r=0.94)was high in WMS method,while the biological repetition correlation(r=0.69)was low.Bray-Curtis distance identified that dissimilarity from biological replicates was larger than that of technical replicates(P<0.001).In addition,dissimilarity and specific taxonomic profiles were observed between 16 S and WMS datasets.Our results imply that homogenization is a necessary step before sample DNA extraction.The sequencers contributed less to taxonomic variation than the library preparation protocols.We developed an empirical Bayes approach that"borrowed information"in calculations and analyzed batch effect parameters using standardized data and prior distributions of(non-)parameters,which may improve population comparability between 16 S and WMS and provide a basis for further application to fusion analysis of published 16 S and microbial datasets.