We have used data derived from hormonal responses in human breast cancer cell lines to develop a panel of 36 genes which can robustly identify patients who will or will not benefit from tamoxifen treatment. These genes had the following characteristics: 1) induction by 17beta-estradiol (E2) in estrogen receptor (ER)-positive breast cancer cell lines in vitro, 2) under-expression in ER-negative breast tumors, and 3) correlation with PR mRNA expression in breast tumors. The average expression of these 36 genes was used as a "risk index" for assessing disease-specific survival in two independent tumor profile datasets of 60 and 67 patients treated with tamoxifen (these data not having been used to initially select the 36 genes), with a high risk group in each dataset defined as those with the bottom 25% of risk index values. This combined biological and informatic analysis is potentially applicable to many other cancer therapeutics.