Fig. 2

Distributions of DNA double-strand breaks in human cells analyzed by GUIDE-seq and iGUIDE. Sequences of suspected Cas9 edited sites associated with either the B2M (a), TRAC5 (b), or VEGFA guideRNAs (c, d). The number of guideRNA mismatches are annotated to the right of the associated sequence, as well as the number of inferred cells sampled, as reported by GUIDE-seq (Gs) or iGUIDE (iG) data. e–g Analysis of the distribution of spontaneous DNA double-strand breaks in cells relative to genomic annotation. Each column shows, from left to right, analysis of sites of dsDNA breaks inferred by iGUIDE and GUIDE-seq. The third column shows sites of lentiviral vector integration in T cells from Fraietta et al. [22] for comparison—HIV favors integration in active transcription units, which is reflected in the integration site preferences [23,24,25]. Rows summarize the relationship of each form of genomic annotation on the human genome to mapped sites. To generate the heat maps, sites are correlated with the density of genomic annotation in intervals along the genome, and co-occurrence summarized as receiver operating characteristic (ROC) curves. Positive associations (> 0.5) are shown by the higher values (red), negative associations (< 0.5) by the lower values (blue). No association (0.5) is shown white. Because the relevant widow size for comparison is unknown, multiple window sizes were tested. Asterisks on each tile compare the statistical significance for comparison to no association. * indicates 0.05 > p > 0.01; ** indicates 0.01 > p > 0.001; *** indicates p < 0.001. e, f: as in (g), but associations are shown relative to epigenetic marks mapped in T cells. In the analysis, 10 Kb chromosomal intervals were used for the comparison