Hierarchical clustering of FLEXIs in different cell lines based on over- and under-represented RBP-binding sites.

<p>FLEXIs containing at least one annotated binding site for any RBP in any of the 4 cell lines (1,261 to 2,404 FLEXIs in different cell lines) were used in this analysis. For hierarchical clustering, subsets of FLEXIs (6 to 1,086) that have binding sites for each of the 47 non-core spliceosomal RBPs with binding sites for ≥30 different FLEXI in a merged TGIRT-seq dataset for the 4 cell lines were extracted for each of the cellular RNA sample types. Contingency tables were then generated by comparing the frequency of binding sites for all 126 RBPs that have an identified binding site in a FLEXI RNA in each of these subsets to those in all FLEXI RNAs. Multiple binding sites for the same RBP in the same FLEXI were counted as one binding site. p-values for over- and under-represented RBP-binding sites calculated by Fisher’s exact test were adjusted by the Benjamini-Hochberg procedure. RBP-binding sites (≥2% abundance) were then key-coded as not significantly different or as significantly over- or under-represented in the tested subset of FLEXIs compared to all FLEXIs (adjusted p≤0.05 calculated by Fisher’s exact test and adjusted by the Benjamini-Hochberg procedure), and the key-coded information was used to construct matrices of the Gower’s distance between the 47 subsets in each of the cellular RNA samples that were used as input for hierarchical clustering by the complete linkage clustering method [<a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011416#pgen.1011416.ref081" target="_blank">81</a>]. The results were displayed as a two-dimensional heat map to identify subsets of FLEXIs showing similar patterns of significantly over- and under-represented RBP-binding sites in each of the cellular RNA samples, with the color scale at the bottom based on -log<sub>10</sub>-transformed adjusted p-values. Significantly co-enriched RBPs are indicated by an X in the heatmap box, and clusters of co-enriched RBPs are delineated in larger boxes. RBP names are color coded by protein function as indicated at the bottom of the Figure.</p> <p>(PDF)</p>