Microbial influences on human health may depend on community-wide characteristics rather than any single taxa. 16S rRNA sequencing can capture the breadth of microbial communities, but analysis of longitudinally-collected, highly-dimensional microbiome data poses significant analytic challenges. I compare the results of clustering via a weighted correlation graph network analysis vs a Dirichlet multinomial modeling approach using a longitudinally collected set of 189 salivary samples from Caucasian children with and without dental decay. Each method has strengths and weaknesses when used to identify oral communities and pivotal microbes which may predispose children to developing dental caries.
