Fig. 5

Multiplet cell-type origins are predicted with high accuracy. a Overview of the cell origin annotation pipeline. First, cells are clustered. Second, marker peaks are identified. Third, multiplets and their k-nearest neighbor cells are used to generate cluster similarity scores. b Example of aggregate cluster profiles for predicting cell origin annotations. Clusters corresponding to cell types observe strong signal for their respective cell types (e.g., cluster 5) while clusters corresponding to multiplets show a mixed profile of cell types (e.g., cluster 13). c Heatmaps of cell origin annotation accuracies for predicting artificial multiplets derived from cells of the specific cell type pairings. Multiplet annotations showed high accuracies for the majority of cell type compositions