Fig. 4

Different configuration of binding and crosslinking across RBPs. a–d Histograms of distances between the cognate motifs of RBPs and the centers of the read coverage peaks (RCRUNCH peak center, in red) or between the motifs and the most frequent read start position in a given peak (RCRUNCH crosslink, in blue). The top 1000 peaks (in the order of their z-score) from one of the available samples for PUM2 (a), PTBP1 (b), RBFOX2 (c), and hnRNPC (d) were extracted. The cognate motif with the highest posterior probability given the RBP’s PWM was determined and the peak was retained only when the posterior was at least 0.3. e For hnRNPC, we carried out the same analysis relative to the Alu-related motif. f Example of two hnRNPC binding peaks located on Alu antisense elements. The library-size-normalized read coverage in the eCLIP IP sample is shown in gray, while the coverage in the corresponding SMI sample is shown in yellow. The fitted Gaussian peaks predicted by RCRUNCH are shown in red, while the distribution of reads starts in this region is depicted with the blue line. The most frequent read start within a coverage peak is chosen by RCRUNCH crosslink as the crosslink position and the corresponding nucleotide is indicated here by the blue arrow. The red arrow links the peak center to the corresponding nucleotide within the Alu antisense element. RPM—reads per million. g Results of CLIP experiment simulations, showing the read coverage profile (full red line), corresponding Gaussian fit (dashed red line), and frequency of crosslinks (blue) with respect to the RBP-specific motif (black, centered on position 0), for low (0.1, left), and high (0.9, right) probability of reverse transcriptase readthrough (\(\rho\))