Human activities impact all ecosystems on Earth, which urges scientists to better understand biodiversity changes across temporal and spatial scales. Environmental DNA (eDNA) metabarcoding is a promising non‐invasive method to assess species composition in a wide range of ecosystems. Yet, this method requires the completeness of a reference database, i.e. a list of DNA sequences attached to each species of the regional pool, which is rarely met. As an alternative, molecular operational taxonomic units (MOTUs) can be extracted as clusters of sequences. However, the extent to which the diversity of MOTUs can predict the diversity of species across spatial scales is unknown. Here, we used 196 samples along the Rhone river (France) for which the reference database is complete to assess whether a blind eDNA approach can reliably predict the ground‐truth number of species at different spatial scales. Using the 12S rDNA teleo primer, we curated and clustered 60 million sequences into MOTUs using a new assembled bioinformatic pipeline. We show that stringent quality filters were necessary to remove artefact noise, notably MOTUs present in a single PCR replicate, which represented 55% of MOTUs (103). Post‐clustering cleaning also removed 19 additional erroneous MOTUs and only discarded one truly present species. We then show that the diversity of retained fish MOTUs accurately predicted the local (α, r = 0.98) and regional (γ) ground‐truth species diversity (67 MOTUs versus 63 species), but also the species dissimilarity between samples (β‐diversity, r = 0.98). This work paves the way towards extending the use of eDNA metabarcoding in community ecology and biogeography despite major gaps in genetic reference databases.

中文翻译：

---

通过聚类环境DNA元条形码序列对整个空间尺度上的脊椎动物分类学多样性进行盲目评估

人类活动影响着地球上所有的生态系统，这促使科学家更好地了解时空尺度上生物多样性的变化。环境DNA（eDNA）元条形码是一种有前途的非侵入性方法，可用于评估各种生态系统中的物种组成。然而，这种方法需要参考数据库的完整性，即附在区域库每个物种上的DNA序列列表，这很少满足。或者，可以将分子操作分类单位（MOTU）提取为序列簇。但是，MOTU的多样性可以预测整个空间尺度上物种多样性的程度尚不清楚。这里，我们在法国罗纳河沿岸使用了196个样本，这些样本的参考数据库已经完成，用于评估盲目eDNA方法能否可靠地预测不同空间尺度上物种的真实数量。使用12S rDNA teleo引物，我们使用新组装的生物信息流水线将6000万个序列编组到MOTU中。我们表明，严格的质量过滤器对于消除伪影噪声是必不可少的，特别是在单个PCR复制中存在的MOTU占MOTU的55％（103）。群集后的清理还删除了19个其他错误的MOTU，仅丢弃了一个真正存在的物种。然后，我们证明了保留鱼类MOTU的多样性准确地预测了当地（α，r = 0.98）和区域（γ）的地面真实物种多样性（67 MOTU对63种），而且样本之间的物种差异（β-多样性，r = 0.98）。尽管遗传参考数据库存在重大差距，但这项工作为在社区生态学和生物地理学领域扩展eDNA元条形码的使用铺平了道路。