Fig. 5

Heatmap of all 5-mers for all available eCLIP datasets. eCLIP datasets are hierarchically clustered by their rank order of 5-mers and regional distribution of high-confidence crosslink sites (tXn) to visualize binding preferences across groups of proteins. Fourteen primary clusters were identified, which are color coded in the dendrogram on the left. K-mers were clustered as described in “Methods.” Additional heatmaps on the left represent (from left to right) (1) regional distribution of thresholded crosslink sites, (2) mean nucleotide composition across top 50 identified k-mers for each dataset, (3) enrichment of the protein in the RNA interactome capture (eRIC [33]), (4) the overlap with orthogonal in vitro methods (i.e., recall), and (5) mean PEKA score across top 50 identified k-mers. RBPs marked on the main heatmap have a recall value > 0.5. A triangle next to the RBP name represents high recall (> 0.8) and a star represents high similarity score (> 0.3). Additional heatmap above the main heatmap represents the nucleotide sequence of each 5-mer and a grayscale heatmap below the main heatmap shows a percentage of eCLIPs, where k-mer ranked among the top 50. Every 20-th 5-mer is labelled on the main heatmap. Yellow fields in grayscale heatmaps indicate missing values