Flow Cytometry Data Analysis Methodologies

Cytometry, especially flow cytometry, is essential for modern immunology (and other biological disciplines). Despite the facts that flow cytometry has been in existence for decades and that automated methods for flow cytometry data analysis have been in the flow cytometry literature since the 1980s, by far manual gating has been the dominant practice. In the Mosmann lab, and with our collaborators, we are working on automated ways to analyze flow cytometry data. Our approaches can be divided into two broad categories: instances in which a population of cells is known in advance to be of interest (as is the case with manual gating), and exploratory analyses. The former is a program GAFF (Gating Assistance For Flow) and the latter is called SWIFT (Scalable Weighted Iterative sampling for Flow Technique). Besides these automated clustering procedures, we are interested in all aspects of automated cytometry data analysis, including the effects of pre-processing and data transformations, data management, data visualization, and, especially, statistical inference. Inference goes beyond finding clusters in a single sample and involves comparisons between samples. Hence, this is the most important aspect of automated data analysis methodology, but it has received perhaps the least attention in the cytometry literature. The cell-by-cell nature of cytometry datasets imparts to them a richness and complexity that is absent in most other types of data, and the fact that the events are not labeled provides additional challenges for inferences across samples.

Three steps in SWIFT to adjust cluster numbers and identify rare populations.