Fig. 4: The design of IPTG-inducible promoters in E. coli.

a The design process of the IPTG-inducible promoter in E. coli. b Schematic representation of the origin sequences, the substitution sequences, and the DeepSEED-designed sequences tested in vivo. We randomly chose the initial promoters from the dataset as the backbone. The −35, −10 elements, spacer length, and the number of lacO sequences as the ‘seed’. The substitution group replaced the corresponding position sequence of the backbone with lacO sites, and the DeepSEED group optimized the flanking sequence of the promoters in the substitution group. c Boxplot showing induced activity and fold-change of the b promoters in vivo. The p-values were determined by a two-tailed unpaired Welch’s t-test, where ns represents not significant. The difference in average induced activity and fold-change of the DeepSEED group and the substitution group are also shown. The promoter activities were normalized by strong constitutive promoter J23100 from the iGEM parts registry. Box plots indicate the median (middle line), 25th, 75th percentile (box) and minimum and maximum values (whiskers). d Two-dimensional images of the induced promoter activity and fold change of the promoters in b. Two wild-used promoters, pLlacO1 and placUV5, are also shown. Each dot represents the average of three biological replicates. e Promoters in the embedding spaces. Each dot represents one natural promoter with certain activities in the dataset. Validation promoters in b are also shown. Source data are provided as a Source Data file.