Figure 4

Application of GOMER to chromatin immunoprecipitation microarray experiments. (a) The contribution to array feature enrichment by binding sites outside the sequence of the array feature. (i) A single protein bound to a single high-affinity site (ii) yields a population of enriched DNA molecules averaging approximately 500 base pairs in length. (iii) Hybridization of the enriched sequences to a DNA microarray results in a signal for those array features that overlap the enriched DNA sequences (N-1 and N). (iv) If the sequence of the array features alone is used to predict binding, enrichment of feature N cannot be accurately predicted. (v) Enrichment can be predicted if flanking sequences are included in the calculation. Binding sites outside the array feature sequence are down-weighted as a function of distance from the array feature boundary. (b) Receiver operating characteristic (ROC) curves for Rap1p enriched versus unenriched features, with features ranked by GOMER scores. GOMER scores were calculated using only the features themselves (left) or the features plus weighted flanking sequences (right). ROC curves for different subsets are indicated by shading under the curve: open reading frame (ORF) features only (light gray); both intergenic and ORF features (medium gray); intergenic features only (dark gray).