Analysis of Microarrays

The microarrays were scanned using an Axon200B scanner, and the images were analyzed with Genepix 5.0. Columns corresponding to the background subtracted intensities and standard deviation of the background were extracted for further analysis. The intensities for the two channels, representing the immunoprecipitated (test) and unenriched (control) samples, were normalized by using the median of each channel to calculate a normalization factor, normalizing all datasets to a single median intensity. The log ratio of the intensity in the test channel to the control channel was calculated. To account for biases in the immunoprecipitation reaction, these log ratios were normalized for each spot by subtracting the average log ratio of each spot across all arrays. The intensities in the test channel were then adjusted to yield this normalized ratio. Finally, an error model was used to calculate significance of enrichment on each chip and to combine data for replicates to obtain a final average ratio and significance of enrichment for each intergenic region. Each intergenic region was assigned to the genes it is most likely to regulate, as described here.

We have included new refinements in our analysis relative to that used in Lee et al. Notably, we have excluded artefactual spots from analysis, selected more reliable probes for normalization and assigned quality metrics to individual arrays to identify low quality experiments.