Probe intensity estimates were log2-transformed and quantile normalized using the Bioconductor package ‘lumi’ (71,72) in the R statistical environment (http://www.r-project.org/). The two batches of samples were first normalized and assessed for quality separately. We then combined and re-normalized the data from all arrays. While assessing the quality of the data (Supplementary data, Figs S3–S5), one sample was identified as a clear outlier and was removed from the study (Supplementary data, Fig. S4). Quality control steps also revealed two batch effects: one segregating samples within the second batch of samples and one segregating the two batches in the combined data set. Because we observed batch effects even after normalization (Supplementary data, Fig. S4), we explicitly regressed them out (as a fixed effect) and the data were then re-normalized.
