Abstract:
|
A key characteristic of microbiome compositional data is its large and complex cross-sample heterogeneity. Appropriately accounting for these “compositional variance components” is critical for a number of common inference tasks, including identifying latent structures, carrying out hypothesis testing on cross-group differences, and modeling dynamics, but is complicated by the key features of microbiome compositional data including high-dimensionality and compositionality. These characteristics incur the need for structural constraints on modeling taxa covariance while maintaining the analytical and computational tractability of the resulting model or method. To address this need, the logistic-tree normal model is introduced as a generative model for microbiome compositions that combines the key features of log-ratio based models and Dirichlet-tree models to allow both structural modeling on taxa covariance and scalable computation. In this talk, I demonstrate how one may utilize the logistic-tree normal model to improve statistical models for identifying subcommunity structures and for characterizing dynamics in microbiome compositions.
|