Supervised classification enables rapid annotation of cell atlases

Single cell molecular profiling technologies are gaining rapid traction, but the manual process by which resulting cell types are typically annotated is labor-intensive and rate-limiting. We describe Garnett, an algorithm and accompanying software for rapidly annotating cell types in scRNA-seq and scATAC-seq datasets, based on an interpretable, hierarchical markup language of cell type-specific genes. Garnett successfully classifies cell types in tissue and whole organism datasets, as well as across species.