Comparison of microbial diversity determined with the same variable tag sequence extracted from two different PCR amplicons. v6_v46_comparison

Background: Deep sequencing of the variable region of 16S rRNA genes has become the predominant tool for studying microbial ecology. As sequencing datasets have accumulated, meta-analysis of sequences obtained with different variable 16S rRNA gene targets and by different sequencing methods has become an intriguing prospect that remains to be evaluated experimentally. Results: Here, we amplified a group of fecal samples using both V4F-V6R and V6F-V6R primer sets, excised the same V6 fragment from the two sets of Illumina sequencing data, and compared the resulting data in terms of the α-diversity, β-diversity, and community structure. Principal component analysis (PCA) comparing the microbial community structures of different datasets, including those with simulated sequencing errors, was very reliable. Procrustes analysis showed a high degree of concordance between the different datasets for both abundance-weighted and binary Jaccard distances (P<0.05), and a meta-analysis of individual datasets resulted in similar conclusions. The Shannon’s diversity index was consistent as well, with comparable values obtained for the different datasets and for the meta-analysis of different datasets. In contrast, richness estimators (OTU and Chao) varied significantly, and the meta-analysis of richness estimators was also biased. The community structures of the two datasets were obviously different and led to significant changes in the biomarkers identified by the LEfSe statistical tool. Conclusions: Our results suggest that beta-diversity analysis and Shannon’s diversity are relatively reliable for meta-analysis, while community structures and biomarkers are less consistent. These results should be useful for future meta-analyses of microbiomes from different data sources.

背景：16S核糖体RNA（16S rRNA）基因可变区的深度测序已成为研究微生物生态学的主流工具。随着测序数据集的不断积累，针对采用不同16S rRNA基因可变区靶标、不同测序方法获得的序列开展的荟萃分析（meta-analysis）已成为极具吸引力的研究方向，但仍有待实验验证。结果：本研究分别采用V4F-V6R与V6F-V6R引物组扩增了一组粪便样本，从两组Illumina测序数据中切取同一V6片段，并从α多样性（α-diversity）、β多样性（β-diversity）与菌群结构三个维度对所得数据进行对比分析。针对不同数据集（含带有模拟测序错误的数据集）的菌群结构开展的主成分分析（PCA）结果具有较高的可靠性。普洛克鲁斯忒斯分析（Procrustes analysis）显示，在加权丰度距离与二元雅卡尔（Jaccard）距离两种指标下，不同数据集之间均呈现高度一致性（P<0.05），且对单个数据集开展的荟萃分析也得到了一致的研究结论。香农多样性指数（Shannon’s diversity index）同样表现出良好的一致性，不同数据集以及不同数据集的荟萃分析所得数值相近。与之相反，丰富度估计量（如操作分类单元（Operational Taxonomic Unit）与Chao指数）的结果存在显著差异，且针对丰富度估计量的荟萃分析也存在偏倚。两组数据集的菌群结构存在显著差异，且通过线性判别分析效应大小（LEfSe）统计工具鉴定得到的生物标志物也发生了明显变化。结论：本研究结果表明，在荟萃分析中，β多样性分析与香农多样性指数具备相对较高的可靠性，而菌群结构与生物标志物的一致性则相对较差。本研究结果可为未来针对不同数据源的微生物组荟萃分析提供重要参考。