Fig. 6: ProMEP identifies beneficial mutations from the gigantic sequence space in the engineering of TadA. | Cell Research

Fig. 6: ProMEP identifies beneficial mutations from the gigantic sequence space in the engineering of TadA.

From: Zero-shot prediction of mutation effects with multimodal deep representation learning guides protein engineering

Fig. 6

a Editing efficiency of TadA variants harboring top 10 single beneficial mutations or deleterious mutations at an endogenous genomic locus (PD1 sg4). Data shown are the comparative fold change in A-to-G conversion efficiency of TadA single mutants relative to ABE1.2. b Accuracy of the identification of both beneficial and deleterious single mutations in TadA by ProMEP. c Base editing efficiency of ABE1.2, ABE8e and TadA multi-site mutants at PD1 sg4 in HEK293T cells. NC negative control. d Base editing efficiency of ABE1.2, ABE8e and TadA multi-site mutants at five endogenous genomic loci in HEK293T cells. e The architecture of TadA-AI-15.8, TadA-AI-15.8-In, ABE8e and ABE9. f The A-to-G conversion efficiency of TadA-AI-15.8, TadA-AI-15.8-In, ABE8e and ABE9 was examined at nine endogenous genomic loci in HEK293T cells. g Average A-to-G conversion efficiency of TadA-AI-15.8, TadA-AI-15.8-In, ABE8e and ABE9 at the 23 target sites. Data are means from three independent experiments. h The C-to-T/G/A conversion efficiency of TadA-AI-15.8, TadA-AI-15.8-In, ABE8e and ABE9 is examined at three endogenous genomic loci in HEK293T cells. NC negative control. i Secondary structure elements of the TadA enzyme are shown. The locations of the substrate-binding loops are indicated in gray, and the mutations in TadA-AI-15.8 and ABE8e are highlighted in red and yellow, respectively. In d, f, h, the heat map represents the average editing percentage derived from three independent experiments. For a, c, data are means ± SD from three independent biological replicates.

Back to article page