A novel approach to modeling immunology data derived from flow cytometry

The flow cytometer (FCM) is a critical measurement device used in a wide variety of fields including Immunology. The technology takes multiple measurements on the surface of biological cells one at time in rapid succession. Variables derived from FCM in the Immunology setting are typically used to find correlations between the flow variables and other immune response measurements or identify changes in mean in a case/control or longitudinal study. The variables are analyzed one at a time using standard statistical tests such as t-tests, ANOVA, and nonparametric tests.;This report will provide an introduction to FCM motivated by examples from the Immunology field and illustrate some of the distributional properties that the variables exhibit. A novel modeling strategy denoted Layered Dirichlet Modeling (LDM), motivated by a generalization of the Dirichlet distribution known as the Nested Dirichlet distribution, will be introduced to model proportions derived from FCM data. The LDM strategy takes into account that the observed values of the variables are constrained and have a hierarchical structure that imposes correlation among the variables. The properties of the LDM testing procedures are explored. A data-driven tree-finding algorithm is proposed to find a hierarchy among FCM cellular subpopulations when the hierarchy is missing or unknown.