Abstract
The subseafloor biosphere is one of the largest ecosystems on Earth, hosting dense populations of microbial cells. Microbial activity is critical to the Earth's geochemical cycles of major elements such as carbon and nitrogen. Despite a growing knowledge of subseafloor microbial diversity, the function and environmental significance of the many uncharacterized lineages remain elusive, along with their importance to the Earth's geochemical cycles. Identification of key microorganisms involved in the cycling of major elements is needed for advancing our understanding of how deep-sea sedimentary microbes influence global climate. Here we use the machine learning technique decision tree to identify which microbial families in deep-sea sediments correlate with geochemical variability in oxygen, nitrate, ammonium, and divalent manganese. We analyzed 1,114 deep-sea sediment samples from 60 sites at water depths ranging from 1,050-10,902 m located along the Arctic Mid-Ocean Ridge (20), the Western North Atlantic Gyre (3), the mid-Atlantic ridge (2), the North-West Pacific Ocean (14), the South China Sea (10), and the South Pacific Ocean (11). Decision tree rule mining revealed deep-sea sedimentary microbial families that likely contribute to subseafloor ecosystem functioning worldwide through four main metabolic pathways: aerobic nitrification, facultative anaerobic heterotrophy, anaerobic ammonium oxidation, and anaerobic heterotrophy. These families may thus be relevant targets for cultivation experiments, representation of functional groups in Earth system models, and monitoring of environmental change in deep-sea sediments caused by environmental perturbations (e.g., deep-sea mining and declining ocean oxygen concentrations).