Analysis of sequences surrounding the site reveals common structural and sequence features.

A) Mean accessibility (unpairedness) of the nucleotides surrounding the cut sites found in the GFP or SOX samples. The accessibility was computed using Localfold [28] on a 300 nt sequence surrounding the cut site (SOX: n = 166, GFP: n = 30). B) Log2 likelihood of each of the 4 nucleotides at position relative to the cut in cut sites shared between the two SOX samples. (Values for the background distribution of the nucleotides were computed from a list of human transcripts) (n = 261). C) Weblogo3 [42] representation of the frequency of each base in the 20 nt surrounding the cut sites found in both SOX samples (n = 267). The position of the cut site is indicated. D) Percentage of sequences with each of the four nucleotides at the cut site (position 0) among the GFP or SOX specific peaks. E) Percentage of sequences surrounding putative cut sites that contain at least one oligo-A stretch within the 6 nt upstream of the cut (for D and E, SOX: n = 267, GFP: n = 53). For all panels in this figure, the shared cut sites were determined based on a scanning window of 4 nt, a multiplicative factor of 4 and a confidence level of 99.99%. All cut sites with the same exact position in both SOX or both GFP samples that were >50 nt away from an annotated transcription start site and had sufficient surrounding sequences within the same annotated exon (300 nt in A, 50 nt in B and 20 nt in C and D) were included in these analyses. The varying number of sequences used for the analyses in the different panels is a result of the requirement for sufficient flanking sequences in the same annotated exon, but as many sequences as possible were analyzed in each case.