Fig. 9

Analyses of a heterogeneous cell population. a Distribution of cell cluster numbers obtained by SCATE for synthetic samples with different cell mixing proportions. GM12878 and K562 cells are mixed at different proportions. For each mixing proportion, 10 synthetic samples are created and analyzed. b An example tSNE plot showing clustering of cells in a synthetic sample. c At each cell mixing proportion, the frequency that each cell type is detected in the 10 synthetic samples is shown. d–f The correlation between the scATAC-seq reconstructed and true bulk regulome in d GM12878, e K562, and f GM12878 and K562 combined for different methods is shown as a function of cell mixing proportion (GM12878 cell percentage). g The peak calling AUC (GM12878 and K562 combined) vs. cell mixing proportion. h The TFBS prediction AUC (GM12878 and K562 combined) vs. cell mixing proportion. i The correlation between the scATAC-seq reconstructed and true bulk differential log-CRE activities is shown as a function of cell mixing proportion. j–l Similar analyses in samples consisting of human CMP and monocyte cells, including j correlation between reconstructed and true bulk log-CRE activities, k peak calling AUC, and l correlation between predicted and true differential log-CRE activities. m–o Similar analyses in samples consisting of mouse thymus and brain cells, including m correlation between reconstructed and true bulk log-CRE activities, k peak calling AUC, and l correlation between predicted and true differential log-CRE activities