Quantifying compositional variability in microbial communities with FAVA

Microbial communities vary across space, time, and individual hosts, generating a need for statistical methods capable of quantifying variability across multiple microbiome samples at once. To understand heterogeneity across microbiome samples from different host individuals, sampling times, spatial locations, or experimental replicates, we present FAVA (F_ST-based Assessment of Variability across vectors of relative Abundances), a framework for characterizing compositional variability across two or more microbiome samples. FAVA quantifies variability across many samples of taxonomic or functional relative abundances in a single index ranging between 0 and 1, equaling 0 when all samples are identical and 1 when each sample is entirely composed of a single taxon (and at least two distinct taxa are present across samples). Its definition relies on the population-genetic statistic F_ST, with samples playing the role of "populations" and taxa playing the role of "alleles." Its mathematical properties allow users to compare datasets with different numbers of samples and taxonomic categories. We introduce extensions that incorporate phylogenetic similarity among taxa and spatial or temporal distances between samples. We demonstrate FAVA in two examples. First, we use FAVA to measure how the taxonomic and functional variability of gastrointestinal microbiomes across individuals from seven ruminant species changes along the gastrointestinal tract. Second, we use FAVA to quantify the increase in temporal variability of gut microbiomes in healthy humans following an antibiotic course and to measure the duration of the antibiotic's influence on temporal microbiome variability. We have implemented this tool in an R package, FAVA, for use in pipelines for the analysis of microbial relative abundances.