2.8 | Null modelling within communities
Like macro-ecological communities, the bacterial microbiome can be shaped by deterministic processes (e.g. selection), stochastic processes (e.g. ecological drift), and dispersal (Adair & Douglas, 2017). If a given community is strongly shaped by selection acting on microbial traits, and microbial traits are phylogenetically conserved, then the phylogenetic structure of this community is expected to deviate from communities assembled through random chance (Webb et al., 2002). Conversely, if a community is strongly influence by ecological drift, then phylogenetic structure of this community is not expected to deviate strongly from null expectations.
To evaluate evidence for the strength of stochastic and deterministic processes in the Sable Island horse microbiome, we first calculated mean nearest taxon distances (MNTDs) using the R package picante (Kembel et al., 2010). MNTD is a measure of the average phylogenetic distance separating every taxon (in this instance ASV) in a community to its nearest neighbour and emphasizes diversity at the tips of a phylogenetic tree. For each horse microbiome, a MNTD null distribution was generated via 9999 randomly assembled communities of ASV richness equal to that of the observed community. Randomized communities were generated by re-shuffling taxa labels and relative abundances across a fixed phylogenetic tree comprising the gamma diversity observed across the entire sample-set. MNTDs were effect size-standardized (MNTDses) relative to the mean and standard deviation of the null distribution for a given community (Stegen, Lin, Konopka, & Fredrickson, 2012). A MNTDses value smaller than -2 or greater than 2 indicate that a community is more phylogenetically clustered or over-dispersed than expected by chance, respectively. While these thresholds have historically been used to make inferences about the relative strength of competition versus environmental filtering (Cavender-Bares, Kozak, Fine, & Kembel, 2009), thought experiments and mixed results from the literature demonstrate such cut-offs are overly simplistic and can lead to a misattribution of pattern to ecological process (Mayfield & Levine, 2010). We instead considered only the magnitude of phylogenetic departure from stochastic expectations (|MNTDses|) in a mixed model inference.