The EPIC data analytics platform for clinical mass cytometry

Abstract

Abstract Single cell technologies, such as high-dimensional cytometry, promise to enable the discovery of immune cell populations that will serve as clinical biomarkers. Recently, we reported the creation of a reference database of the healthy immune system from birth to old age and the development of the EPIC (Extended Poly-dimensional Immunome characterisation) data mining platform (Nature Biotech, accepted). Here we extend the analytics pipeline to facilitate detection of clinically stratifying cell populations in mass cytometry (CyTOF or cytometry by time-of-flight) data. Data structures called immune maps are used to integrate single cell protein expression data (over 40) of multiple samples with clinical metadata and phenotypic information inferred from automated clustering and assisted cell type annotation. Clustering is combined with batch effect correction to reduce technical while preserving biological variations. Subsequent single cell exploratory data analysis and statistical tests help to identify cell populations whose frequencies differ significantly between groups of patients. To gain further insights, users can compare their cytometry data with the healthy reference immunome by performing two types of analysis; (1) mapping to existing clusters helps obtain abundance estimates of immune cell types, (2) reclustering of uploaded data merged with reference immunomes uncovers differences to healthy immune profiles of different ages. We implemented the EPIC pipeline using R Shiny to provide interactive visualisation and an intuitive user interface. We will show examples on how the EPIC platform can characterise cellular diversity in the context of disease and the morphogenesis of the healthy immune system.