Characterization of preferred syntax features of each TF motif. (A) Motif syntax rules modulate TF motif function. For each TF motif type (rows), a linear model was built to predict its activity across all enhancer positions, using as covariates the number of instances, the wild-type TF motif importance and identity, and sequence context features such as the position within the enhancer, the flanking nucleotides, and the presence at close or distal distances to all other TF motifs. The PCC between predicted and observed motif activities is shown with the green color scale on the left. The heatmap shows the contribution of each feature (columns) for each model, colored by the FDR-corrected P-value (red or blue scale depending on positive or negative association, respectively). (B,C) Syntax features associated with GATA (B) or ETS (C) activity. Left: bar plot showing the variance explained by the different types of features (color legend) for each of the linear models. Middle and right: enhancer activity changes (log2 FC to a mutated sequence) after pasting each TF motif in positions with no additional GATA (middle) or ETS (right) in the enhancer, or with additional GATA or ETS at close (≤25 bp) or distal (>25 bp) distances. Number of instances are shown. (D) DeepSTARR-predicted importance scores for pasting a mutant sequence (gray), GATA (blue), or ETS (brown) in a specific position (Chr X: 9,742,091–9,742,339, pos205). Motif sequences pasted are shown. (E) Bar plots with measured enhancer activity (log2) of variants from D.
