ImmPort FLOCK: Automated cell population identification in high dimensional flow cytometry data (42.17)

Abstract

Abstract Historically, investigators have analyzed flow cytometry (FCM) data using manual gating strategies and two-dimensional dot plot visualization to identify and compare specific cell populations between samples. However, this approach is subjective and does not scale well as the number of parameters detectable with different fluorochrome reagents increases. We have developed a software system for automated population discovery in multidimensional FCM data - FLOw Clustering without K (FLOCK), which specifically takes into account the unique features of FCM data, including non-spherical parameter distributions and sparse cell populations. FLOCK includes the following steps: data cleansing, data shrinking, data normalization, informative dimension selection, dense hyper-region identification, histogram partitioning, centroid calculation, clustering, visualization, population statistics calculation, cell ontology mapping. As a result, FLOCK automatically and objectively determines the number of cell populations detectable in a sample and assigns population membership to each cell event, from the multi-parameter flow cytometry data, in the absence of manual gating. Our study of normal human peripheral blood has identified 17 unique B lymphocyte populations based on the expression of IgD, IgG, CD24, CD27, CD38 & B220. Supported by NIH N01AI40076.