To detect genes differentially expressed between treated and control mice, we used the Bioconductor package Limma to fit a linear model for each gene 57. Differential expression analysis produced a top-table, a list of Illumina IDs and their corresponding gene symbols, fold changes (expression level of treated relative to control mice) and p values for the t-test statistic. In order to mitigate the number of false positives but avoid discarding potentially important genes, we used a relaxed statistical significance cut-off value of p < 0.05 (uncorrected) for all analyses in combination with multiple bioinformatics approaches. This technique has been successfully applied in previous studies 7, 34, 42, 44, 48, 66. In all reported analyses, treatment-regulated genes are those differentially expressed at p < 0.05.
