The data are derived from human PBMC from 2 healthy donors, profiled by SingleCell RNAseq (10X platform) at two ICA sites, (Broad/Boston and MtSinai/NYC).

All 4 datasets were integrated using CCA (Butler et al) for construction of the tSNE coordinates. Assignment of each cell to a cell-type follows a tiered process. Cells are first classified and assigned to primary immune lineages classification by fitting their expression programs (multiple regression and naive Bayes optimization) based on ICA gene lists for lineage call, and some manual curation. Cells in each primary lineage are then assigned to “sub-lineages” (e.g. “Naïve CD4+ T”) based on correlation to reference cell sets. These annotations will be further developed to refine the assignment shown, and are expected to evolve over time.

Cells with less than 400 genes present were filtered out of the dataset. The filtered data includes 13,316 cells with an average of 3292 UMIs/cell and an average of 958 present genes/cell. 

PBMC samples sorted by Michal Slyper and Julia Waldman with data processed by Danielle Dionne and Bo Li.