Fig. 5

Systematic evaluation of regulator prediction performance for humans using Cistrome DB ChIP-seq and DNA motif-derived cistromes. a Heatmap showing Lisa’s performance in the analysis of human TF perturbation experiments. Each column represents a TF activation/overexpression or knockdown/out experiment with similar experiment types grouped together. Rows represent the methods based on cistromes from TR ChIP-seq data or imputed from motifs. The upper left red triangles represent the rank of the target TFs based on the analysis of the upregulated gene sets; the lower right blue triangles represent the analysis of downregulated gene sets. The heatmap includes non-redundant human experiments for the same TF. See Additional file 1: Figure S5 for the complete list of human and mouse experiments. b Boxplot showing the target TF rankings comparing Lisa ChIP-seq-based methods and the baseline model based on TF peak counts in gene promoter regions to analyze up- and downregulated gene sets in overexpression/activation (OE) and knockdown/out experiments (KD/KO). c Boxplot showing target TF rankings using Lisa motif-based methods and the baseline model based on motif hits in promoter regions