摘要：The acquisition of biodiversity data is of prime importance to perform an efficient ecosystem management and ensure the sustainable provisioning of marine ecosystem services. Recent efforts are enforcing environmental genomic methodologies, especially environmental DNA (eDNA) metabarcoding, as a way forward to document biodiversity in a faster and cost‐effective way. Currently implemented biomonitoring regulations, such as the Marine Strategy Framework Directive (MSFD), rely on benthic macroinvertebrate indicator taxa to compute biotic indices (BIs) values and derive an ecological quality assessment. Recent work demonstrated that targeting those macroinvertebrates remains challenging, because of various technical and biological biases. Instead, bacterial communities’ profiles depicted by eDNA metabarcoding data have been shown to mirror macroinvertebrate communities’ variation and can be used to accurately predict the BI values using supervised machine learning (SML). Other studies showed that functional profiles, as obtained by metagenomic or metatranscriptomic approaches, vary along pollution gradient in aquatic ecosystems, demonstrating their potential as powerful indicators. Here, a 16S bacterial dataset collected in the vicinity of aquaculture sites, impacted by organic enrichment of the sea bottom, was used to generate two set of features consisting in either OTU profiles (taxonomic turnover) or predicted gene content using the tax4fun tool (functional turnover) that uses taxonomic annotations to infer functional capabilities based on evolutionary models. Both datasets were leading to accurate predictive models, with comparative performance. These results indicate that both taxonomic and functional turnovers of bacterial communities encompass powerful indicators that could be leveraged by marine biomonitoring programs.
