University of Washington - Department of Statistics
In view of the lack of statistical tools to parallel technological advancement in flow cytometry, we propose an automated clustering approach for identifying phenotypically distinct cell populations in flow cytometry data. Our approach is based on finite t mixture models coupled with the Box-Cox transformation, which provides a unified framework to handle outlier identification and data transformation simultaneously. Estimation of the model parameters as well as transformation selection are done via an Expectation-Maximization (EM) algorithm. To demonstrate the proposed methodology, we discuss a cluster analysis on cell population identification in real flow cytometry data. We also give an overview of a Bioconductor software package we have developed to accomplish the goal of such an analysis in flow cytometry.