Fig. 2

Characteristics of EnTDefs. A Overview of the characteristics of 1860 EnTDefs and >5 kb LocDef ranked by F1 score in descending order. F1 score, sensitivity, specificity, number of enhancers, average number of genes per enhancer, average number of enhancers per gene, average proportion of caught TF peaks, average proportion of caught TF peaks outside of 5 kb of TSSs, proportion of genome coverage, and whether the EnTDef was significantly better than the >5 kb LocDef are shown. B The percentage of the 741 EnTDefs with a particular method (four enhancer definition methods, with or without enhancer extension, seven enhancer-target gene link methods) that significantly outperformed the EnTDefs excluding that method only. For simplification, the “nearest_all” additional link method was grouped in enhancer-target gene link method. C Bar plot of average F1 scores for the top 10 EnTDefs plus 5 kb LocDef (blue bars), top 10 EnTDefs (purple bars), nearest TSS method (pink bar), >5 kb LocDef (mustard bar), and 5 kb LocDef used by PE.Approx (yellow bar), by GREAT (dark green bar) and by Fisher’s exact test (green bar). D Distribution of average F1 scores of the top 10 EnTDefs (blue dots) and the nearest_TSS (red dot) among evaluation ChIP-seq datasets or test ChIP-seq datasets. The dashed lines link the same EnTDefs used in the two different ChIP-seq datasets. The p-value of Wilcoxon signed-rank test was shown in the figure