WhichTF robustness analysis.

(a) WhichTF ranking is robust to input region sub-sampling. The top 5 identified TFs are shown for human B-cells, T-cells, heart cells, and brain cells (sample name and the corresponding ENCODE accession ID are shown in the “Sample” column) across the different number of regions in the input files (70%, 80%, 90%, and 100%). The 100% corresponds to the original input file. For the other ones, we subsampled the elements in the BED file based on the SCORE column before applying WhichTF. (b) WhichTF ranking is robust to the lengths of the input region. The top 5 identified TFs are shown for human B-cells, T-cells, heart cells, and brain cells (sample name and the corresponding ENCODE accession ID are shown in the “Sample” column) across different maximum lengths of the regions in the input files (200 bp, 500 bp, 1000 bp, 2000 bp, and original). The ‘original’ correspond to the original input file and for the other ones, we trimmed, if needed, each element in the BED while preserving its midpoint before applying WhichTF. (c) WhichTF ranking is robust to the number of top enriched terms it uses. The top 5 identified TFs are shown for B-cells, T-cells, heart cells, and brain cells (sample name and the corresponding ENCODE accession ID are shown in the “Sample” column) across different numbers of top enriched ontology terms (50, 75, 90, 100 [default], 110, 125, and 150). The ‘100 [default]’ corresponds to the default parameter configuration of WhichTF (see Methods). (XLSX)