Oak Ridge Institute for Science and Education (ORISE) USDA-ARS Research Participation Program, Oak Ridge, TN, USA; USDA-ARS, NEA Bioinformatics, Statistics Group, Beltsville, MD, USA; USDA-ARS, NEA, Beltsville Agricultural Research Center, Animal Biosciences and Biotechnology Laboratory, Beltsville, MD 20705, USA.
USDA-ARS, NEA Bioinformatics, Statistics Group, Beltsville, MD, USA.
Poult Sci. 2022 Aug;101(8):101971. doi: 10.1016/j.psj.2022.101971. Epub 2022 May 21.
The chicken microbiota is often analyzed to address questions about the effects of diet or disease on poultry health. To analyze the microbiota, bioinformatic platforms such as QIIME 2 and mothur are used, which incorporate public taxonomic databases such as Greengenes, the ribosomal database project (RDP), and SILVA to assign taxonomies to bacterial sequences. Many chicken microbiota studies continue to incorporate the Greengenes database, which has not been updated since 2013. To determine whether a choice of database could affect results, this study compared the results of bioinformatic analyses obtained using the Greengenes, RDP, and SILVA databases on a cecal luminal microbiome dataset. The QIIME 2 platform was used to process 16S bacterial sequences and assign taxonomies with Greengenes, RDP, and SILVA. Linear discriminant analysis effect size (LEfSe) was performed, allowing for the comparison of taxonomies considered significantly differentially abundant between the three databases. Some notable differences between databases were observed in results, in particular the ability of SILVA database to classify members of the family Lachnospiraceae into separate genera, while these members remained in one group of unclassified Lachnospiraceae through Greengenes and RDP. LEfSe analyses showed that the SILVA database produced more differentially abundant genera, in large part due to the classification of these separate Lachnospiraceae genera. Additionally, the relative abundance of unclassified Lachnospiraceae in SILVA results was significantly lower than in RDP results. Our results show the choice of taxonomic database can influence the results of a microbiota study at the genus level, potentially affecting the interpretation of the results. The use of the SILVA database is recommended over Greengenes in chicken microbiota studies, as more specific classifications at the genus level may provide more accurate interpretations of changes in the microbiota.
鸡的微生物组通常用于分析饮食或疾病对家禽健康的影响。为了分析微生物组，使用了 QIIME 2 和 mothur 等生物信息学平台，这些平台整合了公共分类数据库，如 Greengenes、核糖体数据库项目（RDP）和 SILVA，以便将细菌序列分配到分类群中。许多鸡的微生物组研究仍在继续使用自 2013 年以来未更新的 Greengenes 数据库。为了确定数据库的选择是否会影响结果，本研究比较了使用 Greengenes、RDP 和 SILVA 数据库对盲肠腔微生物组数据集进行生物信息学分析的结果。使用 QIIME 2 平台处理 16S 细菌序列，并使用 Greengenes、RDP 和 SILVA 分配分类群。进行了线性判别分析效应大小（LEfSe），允许比较三个数据库中被认为显著差异丰度的分类群。在结果中观察到数据库之间存在一些明显的差异，特别是 SILVA 数据库能够将 Lachnospiraceae 家族的成员分类为单独的属，而通过 Greengenes 和 RDP，这些成员仍属于未分类的 Lachnospiraceae 一组。LEfSe 分析表明，SILVA 数据库产生了更多差异丰度的属，这主要是由于这些单独的 Lachnospiraceae 属的分类。此外，SILVA 结果中未分类的 Lachnospiraceae 的相对丰度明显低于 RDP 结果。我们的研究结果表明，在属水平上，分类数据库的选择会影响微生物组研究的结果，这可能会影响结果的解释。在鸡的微生物组研究中，建议使用 SILVA 数据库而不是 Greengenes，因为在属水平上更具体的分类可能会更准确地解释微生物组的变化。
