Direct prediction of regulatory elements from partial data without imputation
(a) Similarity (y-axis) of segmentation produced by different methods on partial data compared with the results obtained from the full data. Results are separated by the number of cell types with missing marks and the number of missing marks in each cell type. Results are further separated by cell types without missing marks (solid boxes) and cell types with missing marks (dashed boxes). Green dashed line indicates median similarity between independent IDEAS runs on the full data set. (b) AUC of precision-recall for predicting peaks of chromatin marks at FDR 0.05 by chromatin states. Results are separated by cell types with the mark being present (solid boxes) and cell types without the mark (dashed boxes).