Extended Data Fig. 2: Features of enhancer activity measured from synthetic random DNA sequences in GP5d cells.

a, Correlation of replicate experiments in GP5d cells. Motif-match log₂ fold change values compared between two replicates from random enhancer experiment (left) and from promoters in binary STARR-seq experiment (right). See legend for Fig. 1c for details of motif naming. b, Motif matches enriched in the oligonucleotides showing the enhancer activity measured from random synthetic DNA (see Methods). c, Comparison of the effect of number of binding sites on enhancer activity from enhancer assay using random synthetic DNA. For each motif, the fold-change compared to the input is shown for one versus two sites. The matching was done separately for each strand so that the background probability of a match was 1 × 10⁻⁵. The black dashed line and the red dotted line represent the expected fold changes if two sites have the same effect as one and if two sites act independently of each other, respectively. d, De novo motif analysis for over-representation of TF motifs within DNA fragments enriched for enhancer activity in GP5d cells from genomic STARR-seq library, from synthetic random DNA library, and from the active transcription factor identification (ATI) assay. The motifs identified by HT-SELEX are shown for comparison. Only the motifs similar to those detected from genomic STARR-seq are highlighted here. Note the ETS-bZIP composite motif enriched in the random STARR-seq data. e-f, Mean of mutual information between 3-mer distributions as a function of distance separating the 3-mers in random enhancers (e) and binary STARR-seq enhancers (f). Pairwise dependency between 3-mer distributions varies with a period of approximately 10 bp and decays as a function of distance separating the k-mers, indicating that most of the dependencies in the enhancers are short-ranged.