Using published microarray data generated from bulk bronchial brushings procured from never and current smokers (GEO series GSE7895), RMA-transformed values for each gene were z-normalized. MetaGene values were then generated by computing the mean z score across all genes in each gene set (GS-1 to GS-19) for each sample. Linear models were built using the stats R package (R v3.2.0), modeling MetaGene expression as a function of donor smoking status and age. For metagenes that were associated with smoking status (FDR q < 0.05), but not age, if the mean current smoker value was greater than or less than the mean never smoker value, the gene set was considered to be up- or down-regulated in current smokers, respectively.