Title: Joint modeling of continuous flow cytometry data with environmental covariates
Authors: Sangwon Hyun - University of Southern California (United States) [presenting]
Jacob Bien - University of Southern California (United States)
Francois Ribalet - University of Washington (United States)
Mattias Cape - University of Washington (United States)
Abstract: Flow cytometry data collected in the ocean can give valuable insight into the composition and dynamics of phytoplankton populations. We present a novel method for modeling time-varying flow cytometry data conditional on a large number of environmental covariates. We develop a novel mixture of multivariate sparse regressions model that can simultaneously estimate and identify the important covariates for each phytoplankton population. The method ties covariates to both the flow cytometry population centers as well as the relative abundances of these populations. The approach involves a lasso-penalized expectation-maximization procedure with additional convex constraints to facilitate interpretation of the estimated model. We apply the method to continuous-time flow cytometry data measured from the ocean, on a ship near Honolulu traveling from warmer, nutrient-sparse subtropical waters to cooler, more productive waters. The method provides a powerful framework for developing a fine-grained understanding of the environmental drivers of phytoplankton populations in the ocean.