Fig. 6: easyCLIP characterizes in vivo RNA–protein interaction landscapes.

a The enrichment of selected RNA categories relative to randomly selected non-RBPs, on a per-read basis. All reads (per million) at the given RNA category were summed for each replicate and the average across replicates is µ. The average µ for randomly selected non-RBPs is µ_non-RBP, and the standard deviation of µ within that group is σ_non-RBP. The size of the markers denotes the enrichment or depletion of a given category as (µ−µ_non-RBP)/σ_non-RBP. P values are two-sided t tests vs the randomly selected non-RBPs (no multiple hypothesis adjustment). b The enrichment of selected RNA categories relative to randomly selected non-RBPs, on a per-protein basis. The calculation was the same as per read except that XLs per protein was used instead. c The number of significantly enriched (FDR < 10⁻⁴) RNAs of the given category for interactions per protein, relative to randomly selected non-RBP controls. d Histograms of cross-links per protein for RBPs and randomly selected non-RBPs; points indicate the left edge of each histogram bin; lines are for easier visualization. (Top) The x-axis is the rate of cross-linking per million cross-linked molecules (i.e., reads per million) to a given RNA converted to a log10 scale, and y-axis is the number of RNAs at that cross-link rate. (Bottom) The x-axis is the rate of cross-linking per 10 billion protein molecules to the given RNA (i.e., reads per million, per 10⁴ proteins), converted to a log10 scale, and the y-axis is the number of RNAs at that cross-link rate. e Protein abundances of WT and mutant RBPs. f Recurrent missense mutations and RBP protein levels (n = 16 WT, n = 18 mutant proteins). For the estimation of difference, the center is the mean and the vertical line indicates the 95% confidence interval. g Volcano plots of changes in easyCLIP reads per RNA induced by recurrent missense mutations in RBPs. DCSH2 and A1CF were increased in E34K above the plotted y values. h Categories of RNAs cross-linked to missense mutant RBPs. i Cross-linking across the PURPL lncRNA is increased per read in KHDRBS2 and A1CF mutants. SF3B1 and FUBP1 are included for comparison. PCBP1 had no cross-linking to PURPL. Data are smoothed and plots have the same y-scale. j Cross-linking across the PURPL lncRNA on a per-protein basis. RNA-seq (Table 8) suggests PURPL lncRNA may be less abundant in cells with R168C KHDRBS2. Together, these results are consistent with a model where affinity for PURPL actually increases (resulting in similar cross-linking per protein), while PURPL RNA levels decrease, resulting in a loss in per-read cross-link frequency.