Table 1.

ER and HER2 Gene Expression Phenotypes

Protein (Positive vs Negative, Charité Data)	Charité (n = 24)	Dana-Farber (Subsampled, n = 24)	Dana-Farber (n = 129)
ER (14 ER+ vs 10 ER−)	414 (FDR = 53%)	2482 ± 1483 (FDR = 11 ± 6%)	9756 (FDR = 1.9%)
HER2 (6 HER2+ vs 18 HER2−)	606 (FDR = 62%)	441 ± 220 (FDR = 88 ± 14%)	654 (FDR = 39%)

The number of differential expressed transcripts between ER+ and ER− as well as HER2+ and HER2− breast cancer samples is presented. The condition p < 0.005 (Welch’s t-test) was used as a threshold for inclusion of a probe set into a gene list. False discovery rates (FDRs) were estimated by a permutation method. Results are shown for the Charité data (n = 24), subsampled Dana-Farber data (n = 24), and the complete Dana-Farber data (n = 129). Subsampling analysis was done by partition of the Dana-Farber cohort in non-overlapping subsets that included the same number of positive and negative tumors as the Charité cohort. Subsequently, mean value and standard deviation over these subsets were calculated and reported.