Fig. 3

Assessment of off-target cleavage events using different prediction methods. A The number of predicted CRISPRScan off-target sites correlated with the number of identified CIRCLE-seq sites (Spearman correlation). Log normalization was used to reduce the range in the number of sites. B In vivo editing scores from the ICE tool for the top predicted off-target sites using CRISPRScan and CIRCLE-seq were not different. Scores were compared using a Mann-Whitney U test. C Editing efficiencies at predicted off-target sites using in vivo scores from Sanger sequencing and mosaicism percentages from Illumina sequencing were correlated (Spearman correlation). D Editing scores obtained in vivo at off-target sites were not correlated with the on-target efficiency of the gRNA. All scatter plots include the Spearman correlations results with the line of best fit