We performed a comprehensive annotation of the clusters using a variety of automated and manual methods. We assigned each cluster a unique mnemonic identifier (e.g., MBDOP1), descriptive name (“midbrain dopaminergic neuron”), major class (e.g., “neuron”), neurotransmitter identity, putative developmental origin, anatomical location, and region (Table S3). We computed enriched genes for each cluster (indicating increased but not unique expression), as well as a probabilistic “trinarization” score, which can be used to determine if a gene is expressed, not expressed, or ambiguous in each cluster (see STAR Methods). We combined enrichment and trinarization scores to discover marker gene sets sufficient to uniquely identify each cluster with high probability (Table S4). Remarkably, we found that 248 (93%) of all clusters were uniquely identifiable with just two genes, while 17 required three genes and none required more than three (although adding more genes increased the robustness of identification).
