Fig. 5: Decoupling the enhancer effect into the combination of individual TFBU effects and optimizing the existing strong enhancer with the TFBU-based framework.
From: Modeling and designing enhancers by introducing and harnessing transcription factor binding units
a Illustration of decoupling the enhancer effect into individual TFBU effects. b The performance of modeling the enhancer activity with different features. The dashed lines denote modeling enhancers with natural language processing models. Error bars denote mean ± s.d. for ten computational replicates. c The measured enhancer activity of optimized CMV enhancers. Error bars denote mean ± s.d. for three biological replicates. d Comparison of enhancer activity between optimized CMV enhancers with one nucleic acid base mutation and more mutations for higher predicted score. The p-value is based on a one-tailed t-test. The box plots show the median (center line), interquartile range between Q1 and Q3 (box limits), and minimum and maximum values (whiskers). Each point denotes the average of three biological replicates. e The importance of each nucleic acid base in CMV enhancers evaluated by the model and important regions as key TFBSs. f The one nucleic acid base mutation caused the highest enhancer activity increase strengthened the core motif signal of RFX5. g The mutation caused a high enhancer activity increase during the greedy optimization strengthened the core motif signal of NFIB. Source data are provided as a Source Data file.
