Analysis Pipeline

Our pipeline of longitudinal mass cytometry data analysis is based on three successive clusterings. Cells sharing similar phenotypes were gathered into cell clusters using SPADE. Then cell clusters displaying similar phenotypes were grouped into phenotypic families. Finally, phenotypic families with similar dynamics were assembled into kinetic families.

The steps of the longitudinal mass cytometry data analysis are displayed. As the first analysis step, single cells from FCS files were grouped into clusters sharing similar phenotype using the SPADE algorithm. Clusters were annotated on the resulting SPADE tree based on the expression of a set of 10 markers, and granulocytes and monocytes-DCs were identified. As the second analysis step, clusters of granulocytes and monocytes-DCs sharing the same categories of marker expression were regrouped into phenotypic families. As the third analysis step, phenotypic families sharing the same kinetic pattern were clustered into kinetic families. Finally, after these three successive clusterings, and as the last analysis step, discriminant analyses were used to determine which kinetic families best distinguished between post-prime and post-boost immune response, and to define the phenotypic signature of each response.