The study compares the performance of five barcoding genes for ecological quality status inference in marine environments. The genes include one bacterial, one foraminiferal and three eukaryotes ones,. Ecological quality status predictive performance of five barcoding genes

Benthic biodiversity monitoring is the standard for ecological impact assessment of anthropogenic activities in marine environments. Recent works showed that high-throughput amplicon sequencing of environmental DNA (eDNA metabarcoding) can overcomes many limitations of the traditional benthic macro-invertebrates inventories approach, by using the molecular data as a surrogate of macrofauna-based biotic indices. Supervised machine learning (SML) proved efficient approach for the building of accurate predictive models from eDNA metabarcoding data, regardless of the taxonomic affiliation of the sequences. Such predictive models are built on a training dataset that include samples from which both macrofauna-based reference biotic indices and an associated molecular dataset is available. However, different barcoding genes may exhibit variable level of accuracy for such endeavor. In this study, we compared five barcoding genes, including one bacterial, one foraminiferal and three eukaryotic genes, all located in the ribosomal Small Sub-Unit (SSU). Our results show that all tested markers are yielding accurate predictive models, and that they all outperform the assessment relying on metazoan assigned sequences from metabarcoding surveys. We further demonstrate how ecological knowledge can be inferred as a side-product of the modelling process.