Prediction of seafloor ecological state using 16S nanopore sequencing.

Anthropogenic stress on benthic habitats, particularly from aquaculture, calls for efficient monitoring of soft-sediment communities. Advancements in Oxford Nanopore Technology, together with e-DNA, offer cost-effective and rapid on-site monitoring of such ecosystems. Studies have demonstrated that Nanopore sequencing provides sufficient precision for predicting ecological state, despite reported challenges with sequencing accuracy. This study aims to predict the seafloor ecological state with both Illumina and Nanopore 16S rRNA data, using a combination of machine learning and feature selection. We analyzed 88 seafloor samples from aquaculture sites located on a north-south gradient along the Norwegian coast. Sequencing methods were evaluated in combination with various bioinformatic approaches in the context of predicting the standard ecological index for aquatic environments, based on macroinvertebrate counting. Our results show predictions from Illumina and Nanopore sequencing platforms are comparable, establishing Nanopore as a feasible alternative to Illumina sequencing. Using a stabilized LASSO regression, the feature set was efficiently optimized from thousands to 40-60 Operational Taxonomic Units (OTUs). This reduced prediction errors to less than half of what was obtained through full feature modeling. These features demonstrated strong predictive accuracy across sequencing technologies, with a high correlation between observed and predicted nEQR values. The Pearson correlation coefficient is 0.98 for Illumina (mean prediction error: ±0.04) and 0.95 for Nanopore data (mean prediction error: ±0.06). This study demonstrates that continual improvements in Nanopore sequencing accuracy, along with optimized feature selection on a broader set of samples, provide a precise and cost-effective monitoring method for marine benthic environments.