The third dataset was taken from the GEO database, GDS accession number GDS534 and reported by Spira, et al. [22]. This study compared bronchial epithelia from humans who were current smokers, never smoked or were former smokers. From this data we selected 20 samples from the current smokers and never smoked groups ("smoking data"). Three samples were omitted because their correlation coefficients with samples in the same group were particularly low, GSM15689 and GSM15692 (current smokers) and GSM15728 (never smoked). The current smoker samples used were GSM15684-GSM15688, GSM15690, GSM15691, GSM15693, GSM15695-GSM15702, and GSM15704-GSM15707; never smoked samples were GSM15718-GSM15727 and GSM15729-GSM15738. These samples were analyzed using MAS5 and were scaled to target intensity of 100. The GEO contained signal and detection p-value. We generated detection calls from the detection p-value using the MAS5 defaults (Present: p < 0.04, Marginal: 0.04 ≤ p < 0.06, Absent: p ≥ 0.06).
