Automated high-dimensional flow cytometric data analysis

Flow cytometry is widely used for single cell interrogation of surface and intracellular protein expression by measuring fluorescence intensity of fluorophore-conjugated reagents. We focus on the recently developed procedure of Pyne et al. (2009, Proceedings of the National Academy of Sciences USA 106, 8519-8524) for automated high- dimensional flow cytometric analysis called FLAME (FLow analysis with Automated Multivariate Estimation). It introduced novel finite mixture models of heavy-tailed and asymmetric distributions to identify and model cell populations in a flow cytometric sample. This approach robustly addresses the complexities of flow data without the need for transformation or projection to lower dimensions. It also addresses the critical task of matching cell populations across samples that enables downstream analysis. It thus facilitates application of flow cytometry to new biological and clinical problems. To facilitate pipelining with standard bioinformatic applications such as high-dimensional visualization, subject classification or outcome prediction, FLAME has been incorporated with the GenePattern package of the Broad Institute. Thereby analysis of flow data can be approached similarly as other genomic platforms. We also consider some new work that proposes a rigorous and robust solution to the registration problem by a multi-level approach that allows us to model and register cell populations simultaneously across a cohort of high-dimensional flow samples. This new approach is called JCM (Joint Clustering and Matching). It enables direct and rigorous comparisons across different time points or phenotypes in a complex biological study as well as for classification of new patient samples in a more clinical setting