tascCODA: Bayesian Tree-Aggregated Analysis of Compositional Amplicon and Single-Cell Data

Abstract

Accurate generative statistical modeling of count data is of critical relevance for the analysis of biological datasets from high-throughput sequencing technologies. Important instances include the modeling of microbiome compositions from amplicon sequencing surveys and the analysis of cell type compositions derived from single-cell RNA sequencing. Microbial and cell type abundance data share remarkably similar statistical features, including their inherent compositionality and a natural hierarchical ordering of the individual components from taxonomic or cell lineage tree information, respectively. To this end, we introduce a Bayesian model for tree-aggregated amplicon and single-cell compositional data analysis (tascCODA) that seamlessly integrates hierarchical information and experimental covariate data into the generative modeling of compositional count data. By combining latent parameters based on the tree structure with spike-and-slab Lasso penalization, tascCODA can determine covariate effects across different levels of the population hierarchy in a data-driven parsimonious way. In the context of differential abundance testing, we validate tascCODA’s excellent performance on a comprehensive set of synthetic benchmark scenarios. Our analyses on human single-cell RNA-seq data from ulcerative colitis patients and amplicon data from patients with irritable bowel syndrome, respectively, identified aggregated cell type and taxon compositional changes that were more predictive and parsimonious than those proposed by other schemes. We posit that tascCODA1 constitutes a valuable addition to the growing statistical toolbox for generative modeling and analysis of compositional changes in microbial or cell population data.

Best practices for single-cell analysis across modalities

“…Univariate statistical models, which analyse change in abundance for each cell type individually, such as Poisson regression or Wilcoxon rank-sum tests, may perceive some cell-type population shifts as statistically sound effects, although they are purely a statistical artefact caused by the compositionality of the data 108 , leading to an elevated FDR. Tests specifically designed for single-cell data that make use of cell-type counts include scDC 109 , scCODA 108 and tascCODA, which can incorporate hierarchical cell-type information 110 .…”

Variational inference for microbiome survey data with application to global ocean data

“…We will explore such modifications in future studies. Moreover, while we chose the Negative Binomial model as base distribution for the most abundant taxa, the variational formulation lends itself to other statistical models for microbial count data, including zero-inflated or hurdle-type extensions of the Negative Binomial model [19] or the Dirichlet-Multinomial model [30, 53]. Finally, in its current state, VI-MIDAS is built on Stan [12] with tailored Python code for optimization, model selection, and analysis.…”

Tree-based differential testing using inferential uncertainty for RNA-Seq

“…There exists another class of methods that fit regression models to compositional data for the covariates using a tree-guided regularization in a maximum likelihood (trac - [Bien et al, 2021]) or Bayesian setting (tascCODA - [Ostner et al, 2021]). For trac, the parsimony of feature selection (where more nodes at greater heights result in a lesser number of total nodes) depends on the weight assigned to regularization during model fitting.…”