From microbial diversity to function; evaluating dimensionality reduction methods

Artificial Intelligence (AI), and more specifically Machine Learning (ML), have become an increasingly prevalent tool in microbial oceanography. The high dimensionality of microbial diversity data from ‘omics observations is highly suitable for ML analysis, with many recent studies showcasing their utility for exploratory ecological feature finding and process prediction. Here, we apply three well-documented dimensionality reduction methods including Principal Coordinate Analysis (PCoA), Self Organizing Maps (SOM), and Weighted Gene Correlation Network Analysis (WGCNA), to near daily 16S rRNA gene amplicon sequencing data from the 2019-2020 MOSAiC International Arctic Drift Expedition. We compare the k-means clustering outputs from these methods to extract functionally distinct seasonal microbial ecotypes in the surface Arctic Ocean. Our results indicate the SOM method outperforms a more traditional PCoA ordination, identifying a greater number of metabolically distinct functional groups. We then investigate the importance of including biological context in dimensionality reduction by comparing functional outputs to a taxa clustering approach using a k-means adapted WGCNA correlation network. Regardless of data input, all 3 methods identified 3-4 recurrent ecotypes with distinct taxonomic and functional cut-offs driven by seasonality, water mass, and substrate turnover. Ultimately, these results reinforce such methodologies as a meaningful translator in the mining of historical amplicon datasets to address modern mechanistic questions and incorporate greater ecotype diversity into mechanistic biogeochemical models.