Johansen, R. | Search bioRxiv

bioRxiv2019

Microbial community-level features linked to divergent carbon flows during early litter decomposition in a constant environment

During plant litter decomposition in soils, carbon has two general fates: return to the atmosphere via microbial respiration or transport into soil where long-term storage may occur. Discovering microbial community features that drive carbon fate from litter decomposition may improve modeling and management of soil carbon. This concept assumes there are features (or underlying processes) that are widespread among disparate communities, and therefore amenable to modeling. We tested this assumption using an epidemiological approach in which two contrasting patterns of carbon flow in laboratory microcosms were delineated as functional states and diverse microbial communities representing each state were compared to discover shared features linked to carbon fate. Microbial communities from 206 soil samples from the southwestern United States were inoculated on plant litter in microcosms, and carbon flow was measured as cumulative carbon dioxide (CO2) and dissolved organic carbon (DOC) after 44 days. Carbon flow varied widely among the microcosms, with a 2-fold range in cumulative CO2 efflux and a 5-fold range in DOC quantity. Bacteria, not fungi, were the strongest drivers of DOC variation. The most significant community-level feature linked to DOC abundance was bacterial richness--the same feature linked to carbon fate in human-gut microbiome studies. This proof-of-principle study under controlled conditions suggests common features driving carbon flow in disparate microbial communities can be identified, motivating further exploration of underlying mechanisms that may influence carbon fate in natural ecosystems.

microbiology↗

bioRxiv2019

Machine learning to predict microbial community functions: An analysis of dissolved organic carbon from litter decomposition.

Microbial communities are ubiquitous and often influence macroscopic properties of the ecosystems they inhabit. However, deciphering the functional relationship between specific microbes and ecosystem properties is an ongoing challenge owing to the complexity of the communities. This challenge can be addressed, in part, by integrating the advances in DNA sequencing technology with computational approaches like machine learning. Although machine learning techniques have been applied to microbiome data, use of these techniques remains rare, and user-friendly platforms to implement such techniques are not widely available. We developed a tool that implements neural network and random forest models to perform regression and feature selection tasks on microbiome data. In this study, we applied the tool to analyze soil microbiome (16S rRNA gene profiles) and dissolved organic carbon (DOC) data from a 44-day plant litter decomposition experiment. The microbiome data includes 1709 total bacterial operational taxonomic units (OTU) from 300+ microcosms. Regression analysis of predicted and actual DOC for a held-out test set of 51 samples yield Pearsons correlation coefficients of .636 and .676 for neural network and random forest approaches, respectively. Important taxa identified by the machine learning techniques are compared to results from a standard tool (indicator species analysis) widely used by microbial ecologists. Of 1709 bacterial taxa, indicator species analysis identified 285 taxa as significant determinants of DOC concentration. Of the top 285 ranked features determined by machine learning methods, a subset of 86 taxa are common to all feature selection techniques. Using this subset of features, prediction results for random permutations of the data set are at least equally accurate compared to predictions determined using the entire feature set. Our results suggest that integration of multiple methods can aid identification of a robust subset of taxa within complex communities that may drive specific functional outcomes of interest.

bioinformatics↗