Biogeographic Patterns in Members of Globally Distributed and Dominant Taxa Found in Port Microbial Communities

Abstract

We conducted a global characterization of the microbial communities of shipping ports to serve as a novel system to investigate microbial biogeography. The community structures of port microbes from marine and freshwater habitats house relatively similar phyla, despite spanning large spatial scales. As part of this project, we collected 1,218 surface water samples from 604 locations across eight countries and three continents to catalogue a total of 20 shipping ports distributed across the East and West Coast of the United States, Europe, and Asia to represent the largest study of port-associated microbial communities to date. Here, we demonstrated the utility of machine learning to leverage this robust system to characterize microbial biogeography by identifying trends in biodiversity across broad spatial scales. We found that for geographic locations sharing similar environmental conditions, subpopulations from the dominant phyla of these habitats (Actinobacteria, Bacteroidetes, Cyanobacteria, and Proteobacteria) can be used to differentiate 20 geographic locations distributed globally. These results suggest that despite the overwhelming diversity within microbial communities, members of the most abundant and ubiquitous microbial groups in the system can be used to differentiate a geospatial location across global spatial scales. Our study provides insight into how microbes are dispersed spatially and robust methods whereby we can interrogate microbial biogeography. IMPORTANCE Microbes are ubiquitous throughout the world and are highly diverse. Characterizing the extent of variation in the microbial diversity across large geographic spatial scales is a challenge yet can reveal a lot about what biogeography can tell us about microbial populations and their behavior. Machine learning approaches have been used mostly to examine the human microbiome and, to some extent, microbial communities from the environment. Here, we display how supervised machine learning approaches can be useful to understand microbial biodiversity and biogeography using microbes from globally distributed shipping ports. Our findings indicate that the members of globally dominant phyla are important for differentiating locations, which reduces the reliance on rare taxa to probe geography. Further, this study displays how global biogeographic patterning of aquatic microbial communities (and other systems) can be assessed through populations of the highly abundant and ubiquitous taxa that dominant the system.

Machine learning applications in microbial ecology, human microbiome studies, and environmental monitoring

“…In practice, ML can perform surprisingly well on datasets that are sampled from and represent messy real-world systems, such as the human body, soil, and water [48] , [49] , [50] and demonstrates superiority over traditional multivariate statistics in analyzing metagenomic data. In addition to these benchmarks, there is an increase in the development of microbiome-specific 'pipelines' that have user-friendly ML implementation and can be accessed through web-interfaces, the statistical compute language R [51] , or Python [52] .…”

“…These studies used hierarchical clustering and non-metric multidimensional scaling (NMDS) to differentiate groups. More recently, SML has been applied to determining the geographic source of an ocean water sample based on the microbial community [50] . Ghannam et al (2020) [50] demonstrated that RF could be used to accurately differentiate the location of sampling of water from 20 different locations.…”

“…More recently, SML has been applied to determining the geographic source of an ocean water sample based on the microbial community [50] . Ghannam et al (2020) [50] demonstrated that RF could be used to accurately differentiate the location of sampling of water from 20 different locations. This study is important in that it shows that SML can be used to identify important trace signals in the microbial community of water that can accurately distinguish between 20 diverse locations from around the world as well as specifically identify the location of collection within locations close in proximity to each other.…”

Machine Learning Predicts the Presence of 2,4,6-Trinitrotoluene in Sediments of a Baltic Sea Munitions Dumpsite Using Microbial Community Compositions

“…To address this issue, higher tax ranks can be used, which are more likely to be found in various areas. Ghannam et al (2020) , for instance, used phyla to differentiate geographic locations on global scale. In our spatially restricted samples the phylum rank also achieved 76.9% mean balanced accuracy, which is still well above coincidence.…”

“…Fingerprints related to community-shaping drivers are revealed by performing unsupervised classification, whereas specific influences can be targeted by the application of supervised machine learning. In microbiological studies, RF has been deployed to predict various geochemical features as well as to detect oil spills ( Smith et al, 2015 ) and to localize the geographic origin of port water across three continents based on dominant bacterial phyla ( Ghannam et al, 2020 ). Moitinho-Silva et al (2017) used RF among other classifiers to separate between sponges of high and low microbial abundance.…”