Fig. 2

Comparison of CLIP methods against RBNS. Heatmap shows the rank order of k-mer PEKA scores for each CLIP dataset, and of RBNS z-scores. The scale on the top right shows these values spanning from 1 to 1024. K-mers selected for this heatmap (n=230) ranked among the top 10 in eCLIP, PAR-CLIP, iCLIP, or RBNS for any of the RBPs shown. K-mers were clustered based on their sequence and ordered using hierarchical clustering of median dataset ranks within clusters. On the left, prominent k-mer sequence features are highlighted across clusters. Below, the recall heatmap shows how much the top motifs in the dataset correspond with the RBNS data. The third heatmap shows the percentage of tXn derived from different transcript regions, and the fourth heatmap shows the percentage of tXn found in repeat sequences. eCLIP experiments in bold are in HepG2, others in K562 cell line. In case two eCLIP datasets for the same RBP were available, the one with higher recall is shown