Strengths and biases of morphological and NGS data on eukaryotic microbiome characterizations

Molecular surveys of eukaryotic microbial communities, employing next-generation sequencing (NGS) techniques, are rapidly supplanting traditional morphological approaches due to their larger data output and reduced bench working time. Although several biases are known to hamper these analyses, it is not yet entirely clear how to quantify them. Here we directly compare morphological and NGS data obtained from the same samples, in an effort to characterize ciliate communities from freshwater sediments in different environments. We show how in silico processing affects the final outcome, suggesting that stringent filtering provides more consistent results. We determine the abundance distribution of ciliate communities, showing that a small fraction of extremely abundant taxa dominate read counts. At the same time, we advance reasons to believe that errors affecting NGS abundances may be significant enough to blur some biological realities. We confirmed that the NGS approach detects far more taxa than morphological inspections, but that the difference varies among taxonomic groups. Finally, we hypothesize that the two datasets actually correspond to different interpretations of “diversity”, and consequently that neither is entirely superior to the other when investigating environmental protists.

利用下一代测序（next-generation sequencing, NGS）技术开展的真核微生物群落分子调查，因其数据产出量更高、实验室操作耗时更短，正快速取代传统形态学研究方法。尽管已知存在若干偏倚会干扰此类分析，但目前仍未完全明确如何对这些偏倚进行量化。本研究直接比对了同一样本的形态学与NGS数据，旨在解析不同环境下淡水沉积物中的纤毛虫群落。我们阐明了计算机处理流程对最终分析结果的影响机制，提示严格过滤步骤可获得更具一致性的分析结果。本研究明确了纤毛虫群落的丰度分布特征，发现极小一部分丰度极高的分类群主导了测序读段计数结果。与此同时，我们提出佐证理由，认为影响NGS丰度数据的误差可能足够显著，足以掩盖部分生物学真实状态。我们证实，NGS方法可检测到远多于形态学观测的分类群，但两类方法的检测差异因分类群类群而异。最后，我们提出假说：两种数据集实际上对应对‘多样性’的不同解读，因此在调查环境原生生物时，二者并无绝对的优劣之分。