Bayesian Hierarchical Model for Single-Cell Assays

• Activity Number: 541
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Lynn Lin*+ and Greg Finak and Raphael Gottardo
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: single-cell assay ; Immunology ; multivariate hypothesis ; Bayesian hierarchical model

Abstract

In immunological studies, the characterization of cell subsets is crucial to decipher system level biological changes. However, many functionally distinct cell subsets can only be measured accurately using single-cell assays. For this reason, a large number of studies rely on analyzing single-cell assays. A common statistical problem is, for every marker combination, to identify subjects for whom that combination of marker are significantly different between two experimental conditions, e.g. before/after vaccination, where the difference is defined as the proportion of cells expressing that marker combination in the cell subsets of interest. We present here a Bayesian hierarchical framework for such multiple hypotheses testing based on a structured Dirichlet-Multinomial model, to testing multivariate differential expression across multiple marker combinations. Our model allows the inference to be subject specific, as is typically required when accessing vaccine responses. We illustrate this multivariate approach through both simulations and single-cell data, we also compared our method against state-of-the-art alternatives.