Phylogenetic factorization of compositional data

1 Marker gene sequencing of microbial communities has generated big datasets 2 of microbial relative abundances varying across environmental conditions, sam3 ple sites and treatments. These data often come with putative phylogenies, 4 providing unique opportunities to investigate how shared evolutionary history 5 a ects microbial abundance patterns. Here, we present a method to identify the 6 phylogenetic factors driving patterns in microbial community composition. We 7 use the method, phylofactorization , to re-analyze datasets from human body 8 and soil microbial communities, demonstrating how phylofactorization can be 9 a dimensionality-reducing tool, an ordination-visualization tool, and also mass10 produce inferences on the edges in the phylogeny in which meaningful di erences 11 arose. 12