Fig. 2

Design of a synthetic compensatory path. In order to generate a set of enhancers with decreasing similarity to MSE2, first a sequence level model of gene regulation (SLM) is trained. This model predicts expression from the underlying structure of binding sites. a The binding structure of the minimal stripe 2 element is shown. The height of bars represents the percent of the maximum log-likelihood score of a motif at each position. Putative activators are plotted on the positive axis and putative repressors on the negative axis. A subset of motifs are shown. Binding sites for all factors considered in this work are included in Additional file 1: Figures S5-S15. b Expression levels of enhancers are used to train an SLM. The observed expression levels (black) and model fit (red) are shown along the AP axis. c The binding structure of a putative synthetic stripe 2 element designed in silico using the SLM in panels A and A is shown, plotted as in panel A. d The predicted expression of the synthetic enhancer in C is shown plotted as in panel B. e A synthetic compensatory path is found that will transform MSE2 into the synthetic element while preserving predicted expression of a stripe. The root mean squared (rms) differences between predicted expression and MSE2 expression is shown as a function of number of sequence mutations. Each step along the x-axis represents a single nucleotide change to the previous sequence. A set of these elements (open circles) are selected for validation in vivo