Deciphering eukaryotic cis-regulatory logic with 100 million random promoters

Deciphering cis-regulation, the code by which transcription factors (TFs) interpret regulatory DNA sequence to control gene expression levels, is a long-standing challenge. Previous studies of native or engineered sequences have remained limited in scale. Here, we use random sequences as an alternative, allowing us to measure the expression output of over 100 million synthetic yeast promoters. Random sequences yield a broad range of reproducible expression levels, indicating that the fortuitous binding sites in random DNA are functional. From these data we learn models of transcriptional regulation that predict over 94% of the expression driven from independent test data and nearly 89% from sequences from yeast promoters. These models allow us to characterize the activity of TFs and their interactions with chromatin, and help refine cis-regulatory motifs. We find that strand, position, and helical face preferences of TFs are widespread and depend on interactions with neighboring chromatin. Such massive-throughput regulatory assays of random DNA provide the diverse examples necessary to learn complex models of cis-regulatory logic.