FROM THERMAL SPRINGS TO SUBWAY BENCHES: EXPLORING THE DIVERSITY OF CARBON MONOXIDE DEHYDROGENASES THROUGH METAGENOMES, PHYLOGENETICS, AND MACHINE LEARNING

Carbon monoxide is well known as a toxic gas but can also be an important input and intermediary for microbial metabolisms. Carbon monoxide dehydrogenases (CODHs) serve as key enzyme complexes for a variety of microbial carbon monoxide (CO) utilization pathways. Such pathways include the Wood-Ljungdahl pathway, which is important in methanogenesis and acetogenesis, metal and sulfate reduction pathways, hydrogen production, and others. The CODH enzymes allow microbes to turn the traditionally toxic waste gas of CO into a useful carbon and energy source. Despite the flexibility of CODH enzymes, the use of carbon monoxide is still believed to be a fringe metabolism. Here we seek to expand the known diversity, distribution, and phylogeny of CODH catalytic subunit proteins by searching an expansive dataset of over 50,000 metagenome assembled genomes. Our work has shown that this dataset contains 5,426 putative CODH protein sequences found within 4,001 metagenome assembled genomes. Despite the considerable expansion of the known set of CODH sequences, our phylogenetic analysis has validated the protein\u27s previously established phylogeny while showing a wider environmental and taxonomic distribution of CODHs. Often considered to be found primarily in areas with high levels of CO, CODHs are typically associated with thermal and extremophiles. In addition to the expected high temperature environments, CODHs were found in metagenomes from diverse environments from soils to subway benches, and in phyla ranging from archaeal Euryarchaeota to bacterial Actinobacterota. We also have constructed a machine learning model to extract functional predictions and information using a sequence-only method to predict gene ontologies (GO-terms) associated with CODH function. While our model can achieve accurate prediction of GO-terms, our work has shown some of the current limitations in the approach. This study reveals CODHs to be a more diverse and ubiquitous enzyme than previously anticipated. Despite tripling the number of sequences in the phylogeny, we provide strong support for the previously established clades and report no new clades. This work has also identified some key areas for experimental follow up regarding the importance of carbon monoxide and CODH genes in many environments