PMC

A mixture model is comprised of multiple simpler distributions added together to describe a more complex distribution. In this way, with a sufficient number of simple components (typically multivariate normal distributions), arbitrarily complex distributions can be fitted to flow cytometry data

Traditional small, tight lymphocyte gates will exclude most dividing lymphocytes in a CFSE assay. Viable (amine low) CD3+CD4+ and CD3+CD8+ lymphocyte clusters are plotted to illustrate that proliferating CFSE-low cells (red) have highly atypical scatter characteristics and are likely to be excluded using standard “lymphocyte” gating strategies that are effective for non-proliferating lymphocytes (orange)

Mixture models reveal potential false negatives and false positives in an ICS assay. The left set of panels show ICS-positive events (small arrows) that are CD3+CD4+ or CD3+CD8+ but have atypical FSC/SSC characteristics possibly due to their state of activation that will be false negatives with a standard gating strategy. The middle panel shows a diagonal streak (large arrow) from non-specific binding due to dead cells. As this is classified as a separate cluster, it is trivial to eliminate such false positives with model-based analysis, but is otherwise extremely challenging with manual analysis. Right set of panels shows the relative frequency of ICS positive CD4+ and CD8+ cells obtained with manual analysis and automated clustering

Identification of tetramer-binding CD4-CD8+ cells with model-based analysis. Data are from an unpublished immune monitoring study of a multi-peptide vaccination trial designed for HLA-A2 patients with biochemical relapse after prostatectomy. The figure shows the frequency of cells binding to HLA-A2 tetramers (relative to CD4-CD8+ cells) constructed using an epitope derived from the prostate specific membrane antigen PSMA 711-719 ALFDIESKV before (top panel) and after (bottom panel) four vaccinations in one patient. Cells in both panels were expanded in vitro in the presence of the specific peptide and interleukins before staining. Background events are gray, CD4-CD8+ events are yellow and events from CD4-CD8+ tetramer+ are brown. Clusters were defined against the mean (μ) and standard deviations (SD) of all events from the before vaccination panel as follows: clusters were designated as CD4− if the mean CD4 of the cluster was less than μ-SD, CD8+ if the mean CD8 of the cluster was greater than μ + SD, and tetramer+ if the mean tetramer of the cluster was greater than μ + 3 SD