Description
ChIPOTle is a Microsoft Excel add-in Macro that analyzes yeast ChIP-chip data generated on whole-genome tiled arrays.

Introduction
In contrast to mRNA microarray experiments, in which each arrayed element usually measures the abundance of one mRNA species, in ChIP-chip experiments each element measures the abundance of a population of fragments of assorted lengths due to chromatin shearing. Therefore, arrayed elements representing genomic regions 1- to 2-kb downstream or upstream from the binding site will also detect enrichment. This effect produces a peak over several arrayed elements containing genomically adjacent DNA. This is non-random behavior that is not expected from spuriously high ratio measurements. ChIPOTle takes advantage of this fact and uses it as an independent confirmation of enrichment for a given genomic region.

ChIPOTle works by determining a sliding window average across each chromosome. A window of selected size (default 1 kb) is slid across a region or chromosome, and the average log2 ratio of any arrayed elements that fall within that window is determined. The window is moved downstream by the step size (default 0.25 kb), and then the calculation is repeated iteratively for the whole chromosome. This sliding average will identify binding sites as peaks. The height of peaks caused by spuriously high ratios will be reduced, since the probability of a neighboring genomic element also having a high ratio is low. ChIPOTle also defines a confidence value for each peak based on the number of independent arrayed elements used to construct the peak.

The utility of this approach is that it does not depend on the absolute number of targets, but on the density of their distribution. It is appropriate for detecting any number of targets that are distributed with a frequency less than approximately three times the average sheared chromatin size. For example, if the average sheared chromatin size were 1 kb, this method would be useful for the detection of any protein predicted to be spaced at intervals of at least 3 kb. A drawback to this approach is that it requires high-resolution tiling arrays.