Flow Cytometry Data Analysis: Comparing Large Multivariate Data Sets Using Classification Trees

This paper describes a method to compare flow cytometry data sets, which typically contain 50,000 six-parameter measurements each. By this method, the data points in two such data sets are divided into subpopulations using a binary classification tree generated from the data. The test is then used to establish the homogeneity of the two data sets based on how their data are distributed across these subpopulations. Preliminary results indicate that this comparison method is sufficiently sensitive to detect differences between flow cytometry data sets that are too subtle for human investigators to notice.