Exon array analysis.
The Affymetrix Power Tools (APT) suite of programs was obtained from http://www.affymetrix.com/support/developer/powertools/index.affx. Exon (probeset) and gene-level signal estimates were derived from the CEL files by RMA–sketch normalization as a method in the apt-probeset-summarize program. To determine if the signal intensity for a given probeset is above the expected level of background noise, we utilized the DABG (detection above background) quantification method available in the apt-probeset-summarize program as part of Affymetrix Power Tools (APT). Briefly, DABG compared the signal for each probe to a background distribution of signals from anti-genomic probes with the same GC content. The DABG algorithm generated a p-value representing the probability that the signal intensity of a given probe was part of the background distribution. We considered a probeset with a DABG p-value lower than 0.05 as detected above background. The statistic thCNS-SCns,ESC = (μhCNS-SCns − μESC) / sqrt (((nhCNS-SCns − 1)σ2 hCNS-SCns + (nESC − 1)σ2 ESC)(nhCNS-SCns + nESC)) / ((nhCNS-SCnsnESC) (nhCNS-SCns + nESC − 2))), where nhCNS-SCns and nESC were the number of replicates, μhCNS-SCns and μESC were the mean, and σ2 hCNS-SCns and σ2 ESC were the variances of the expression values for the two datasets used to represent the differential enrichment of a gene using gene-level estimates in hCNS-SCns relative to hESCs. Multiple hypothesis testing was corrected by controlling for the false discovery rate (Benjamini-Hochberg).