Recent advances in flow cytometry (FCM) allow for distinguishing between increasing
numbers of cell subpopulations in the same tissue sample. However, standard one-dimensional histograms
and two-dimensional plots are insufficient to capture fully the multi-dimensional nature of FCM data even when
attention is restricted to just one tissue sample. These difficulties compound in complex study designs.
Alternative methods for analysis and visualization are necessary to harness the full power of FCM. Our
overarching goal is to investigate the use of advanced computational and statistical methods that take
advantage of the high dimensionality of FCM data, and ultimately to develop iFlowCyt, an integrated,
interactive, internet-based software tool that implements these methods with a friendly user interface.
Self-organizing maps for automated gating
We are examining the use of self-organizing maps (SOMs) for exploring and clustering subpopulations of
cells within tissue samples, and will extend to multiple tissue samples from individual patients. SOM can
map multi-dimensional FCM data to 2 dimensions through the use of self-organizing neural networks and
group similar cells together so that human experts can inspect and discover cell subpopulations across
multiple channels simultaneously.
Statistical graphics for interpretion of gating results
We are assessing the use of alternative graphical methods for representing multi-dimensional data across
individual tissue samples.
If a sample is divided into multiple cell sub-populations, we use the
wind-rose chart (WRC) to display simultaneously the percent of cells in each sub-population. However, as the
number of sub-populations increases, the ability to differentiate the segments in one WRC decreases. Hence
we will also examine the use of other methods (e.g., heat maps, mosaics, etc.) to display these data, both as
percentages in sub-populations as well as absolute counts of cells in sub-groups.
Statistical methods for interpretation of gating results
We are investigating the use of compositional data analysis (CDA) for comparing cell subpopulations between
groups of tissue samples. Since FCM data can be expressed as absolute counts of cells in various subpopulations
as well as proportions, we need to examine tools that can address both types. We will examine
CDA to examine the proportions of cells in various sub-populations, to estimate central tendency and variability
in groups of samples, and to compare between groups of samples. We also investigate the use of other
multivariate methods to examine simultaneously the absolute counts of cells in various sub-populations.
iFlowCyt is our research prototype that is under development. It is an integrated, interactive, internet-based set of software
tools that implement the studied methods. You are more than welcome to test our current release IFCSoft 0.4 and let us know if you have any questions or suggestions!