Statistical Features of Yeast’s Transcriptional Regulatory Code

Using a newly developed algorithm, MotEvo, on a combination of comprehensive ChIP-on-chip binding data, the genomes of 5 related species of yeast, and a collection of experimentally known binding sites we have produced an extensive genome-wide annotation of regulatory sites in the yeast genome containing over 12, 000 high confidence binding sites for 72 regulatory factors. Here we study the large-scale properties of this regulatory code. We find that, in contrast to previously reported results, both the number of binding sites per factor and the number of binding sites per regulatory region follow exponential distributions. Divergently transcribed regions contain twice as many sites as unidirectionally transcribed regions, suggesting that each site regulates only one of the two downstream genes. For eleven factors we find that binding sites show a strong strand preference. The distribution of all binding sites shows a sharp peak at about 100 bases upstream of transcription start and falls exactly exponentially both upstream and downstream of this location. Analysis of the distance distributions of individual factors shows that they can be divided into a number of subclasses that show similar distance distributions of their binding sites.