Direct prediction of regulatory elements from partial data without imputation
(a) Concordance of cell type distances inferred from partial data by different methods compared with the distances inferred from full data segmentation. Concordance is measured by Spearman correlation between two distance vectors for each cell type. Red: joint segmentation; orange: two-step segmentation; green: imputation; blue: concatenation; grey: common mark. (b) Distribution of segmentation-based distances between cell types with and without missing data. Dotted boxes: distances between cell types with missing data; dashed boxes: distances between cell types with missing data and cell types without missing data; solid boxes: distances between cell types without missing data. Green dashed line: median distance between cell types obtained from full data segmentations.