Fig. 6: Functional significance of highly brain-specific unannotated 3’UTRs.

a Violin plot comparing the RBP-binding density across known 3’UTRs, predicted 3’UTRs and predicted non-3’UTRs, categorised according to their tissue-specificity. Only regions with RBP enrichment score greater than zero were displayed. Box plots show the median value (middle line), 25th and 75th percentile (box), and 1.5 times the interquartile range (whiskers). See Supplementary Table 9 for sample size. b Density distributions comparing the 'constrained non-conserved' (CNC) scores between known 3’UTRs and predicted 3’UTRs, categorised according to their tissue-specificity. See Supplementary Table 10 for sample size. p: p value of comparison calculated using two-sided Wilcoxon rank-sum test; es: Wilcoxon effect size (r). c GO terms enriched amongst the list of genes associated with highly brain-specific unannotated 3’UTRs. CC cellular component. d SynGO terms over-represented in genes associated with highly brain-specific 3’UTRs. CC cellular component, BP biological process. e Genome browser view of the APP locus, showing unannotated 3’UTR detected downstream of APP in the hypothalamus, RNA-seq expression in bigwig format (.bw), poly(A) sites from the poly(A) atlas, and the human genome mappability scores from UCSC (Umap). Umap S50 and S100: Single-read mappability for 50- and 100-mers; Umap M50 and M100: Multi-read mappability for 50- and 100-mers.