Fig. 3: On- and off-target activities of SuperFi base editors in mammalian cells.

a, c Base editing activities of CBE, ABE7, and ABE8e variants were assessed in the BEAR assay with a 20 matching and c 33 mismatching sgRNAs. The same data are shown by individual targets in Supplementary Figs. 7a and 9a, respectively. b, d Base editing activities of ABE8e variants on endogenous b on-target and d off-target sites as measured by NGS. Data are also presented in Supplementary Figs. 7b and 9b, respectively. b The editing efficiency (16 genomic sites) of adenines inside (A4-A8, n = 28) and outside the editing window (bystander editing, n = 41) is shown side by side, separated with a dashed line. Regarding base editing, the numbering of the edit positions follows the convention laid in the literature, i.e., 5’ to 3’ direction in the non-targeted strand of the target DNA. c The fidelity of ABE8e variants was assessed in the BEAR assay with 2 matching sgRNAs and 33 sgRNAs mismatching in one (MM1; n = 18) or in two to three positions (MM2-3; n = 15). The relative activity of all base editors (off-target/on-target ratio) for all adenines is plotted separately for all MM1 and all MM2-3 sgRNAs. d The relative activity (off-target/on-target ratio) of ABE variants is shown for all adenines (n = 43) of 17 genomic off-target sites. For on-target adenine value, we selected the adenine which was edited by ABE7 (for EMX1 by ABE8e) at the highest level for each on-target sequence. a–d Results are presented on a scatter dot plot, the median and interquartile ranges are shown; data points are plotted as open circles representing the means of triplicates. Summary of target and primer sequences, BEAR, NGS data and allele tables are reported in Supplementary Data 1–3 and 7. Statistical significance was assessed by RM one-way ANOVA, statistical details and p-values are available in Methods and in Supplementary Data 6 (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).