Srinivas Meghana, Walsh Calum J, Crispie Fiona, O'Sullivan Orla, Cotter Paul D, van Sinderen Douwe, Kenny John G
Teagasc Food Research Centre, Fermoy, Co. Cork, Ireland.
School of Microbiology, University College Cork, Cork, Ireland.
Sci Rep. 2025 Jan 22;15(1):2822. doi: 10.1038/s41598-024-83410-7.
Rapid advancements in long-read sequencing have facilitated species-level microbial profiling through full-length 16S rRNA sequencing (~ 1500 bp), and more notably, by the newer 16S-ITS-23S ribosomal RNA operon (RRN) sequencing (~ 4500 bp). RRN sequencing is emerging as a superior method for species resolution, exceeding the capabilities of short-read and full-length 16S rRNA sequencing. However, being in its early stages of development, RRN sequencing has several underexplored or understudied elements, highlighting the need for a critical and thorough examination of its methodologies. Key areas that require detailed analysis include understanding how primer pairs, sequencing platforms, and classifiers and databases affect the accuracy of species resolution achieved through RRN sequencing. Our study addresses these gaps by evaluating the effect of primer pairs using four RRN primer combinations, and that of sequencing platforms by employing PacBio and Oxford Nanopore Technologies (ONT) systems. Furthermore, two classification methods (Minimap2 and OTU clustering), in combination with four RRN reference databases (MIrROR, rrnDB, and two versions of GROND) were compared to identify consistent and accurate classification methods with RRN sequencing. Here we demonstrate that RRN primer pair choice and sequencing platform do not substantially bias taxonomic profiles for most of the tested mock communities, while classification methods significantly impact the accuracy of species-level assignments. Of the classification methods tested, Minimap2 classifier in combination with the GROND database most consistently provided accurate species-level classification across the communities tested, irrespective of sequencing platform.
长读长测序技术的快速发展，通过全长16S rRNA测序（约1500 bp），更显著地是通过更新的16S-ITS-23S核糖体RNA操纵子（RRN）测序（约4500 bp），促进了物种水平的微生物分析。RRN测序正成为一种更优越的物种分辨率方法，超越了短读长和全长16S rRNA测序的能力。然而，由于处于发展初期，RRN测序有几个未充分探索或研究不足的方面，这凸显了对其方法进行批判性和全面审视的必要性。需要详细分析的关键领域包括了解引物对、测序平台、分类器和数据库如何影响通过RRN测序实现的物种分辨率的准确性。我们的研究通过使用四种RRN引物组合评估引物对的影响，以及通过采用PacBio和牛津纳米孔技术（ONT）系统评估测序平台的影响，来填补这些空白。此外，还比较了两种分类方法（Minimap2和OTU聚类）与四个RRN参考数据库（MIrROR、rrnDB和两个版本的GROND），以确定与RRN测序一致且准确的分类方法。我们在此证明，对于大多数测试的模拟群落，RRN引物对的选择和测序平台不会对分类学图谱产生实质性偏差，而分类方法会显著影响物种水平分类的准确性。在所测试的分类方法中，Minimap2分类器与GROND数据库相结合，无论测序平台如何，在所有测试群落中最一致地提供了准确的物种水平分类。
