RT Journal Article
SR Electronic
T1 Uncovering the drivers of animal-host microbiotas with joint distribution modeling
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 137943
DO 10.1101/137943
A1 Johannes R. Björk
A1 Francis KC. Hui
A1 Robert B. O’Hara
A1 Jose M. Montoya
YR 2017
UL http://biorxiv.org/content/early/2017/05/15/137943.abstract
AB Background In addition to the processes structuring free-living communities, host-associated microbial communities (i.e., microbiotas) are directly or indirectly shaped by the host. Therefore, microbiota data have a hierarchical structure where samples are nested under one or several variables representing host-specific features. In addition, microbiota data are often collected across multiple levels of biological organization. Current statistical methods do not accommodate this hierarchical data structure, and therefore cannot explicitly account for the effects of host-specific features on structuring the microbiota.Methods We introduce a unifying model-based framework developed specifically for analyzing host-microbiota data spanning multiple levels of biological organization. While we chose to discern among the effects of host species identity, host phylogeny, and host traits in structuring the microbiota, the presented framework can straightforwardly accommodate any recorded data that includes host-specific features. Other key components of our modeling framework are the powerful yet familiar outputs: (i) model-based ordination to visualize the main patterns in the data, co-occurrence networks to visualize microbe-to-microbe associations, and (iii) variance partitioning to asses the explanatory power of the included host-specific features and how influential these are in structuring the microbiota.Results The developed framework was applied to published data on marine sponge-microbiota. We found that a series of host traits that are likely phylogenetically conserved underpinned differences in both abundance and species richness among sites. When controlling for these differences, microbiota composition among sites was confounded by numerous site and host-specific features. At the host level, host traits always emerged as the prominent host-specific feature structuring the microbiota.Conclusions The proposed framework can readily be applied to a wide range of microbiota systems spanning multiple levels of biological organization, allowing researchers to systematically tease apart the relative importance of recorded and/or measured host-specific features in structuring the microbiota. The study of free-living species communities have significantly benefited from the increase in model-based approaches. We believe that it is time for research on host-microbiota to leverage the strengths of a unifying model-based framework.