Expression profiling of peripheral blood gene expression of women with hereditary breast cancer and controls (set 1)

We obtained peripheral blood samples for women from Utah (USA) and Ontario (Canada) who had a family history of breast cancer (or did not), who carried a BRCA1/2 mutation (or did not), and who had developed breast cancer (or had not). We classified the women into two groups: [1] those who had a family history of breast cancer (irrespective of BRCA1/2 mutation status) and had developed an early-onset breast tumor and [2] those who had a family history of breast cancer but had not developed a breast tumor or who did not have a family history of breast cancer (some of whom had developed sporadic breast cancer and and others who had not). We then used machine-learning methods to assess how well we could classify these women into either group. The Utah samples served as a training set, and the Ontario samples served as an independent validation set from a geographically distinct population. For the Utah cohort, there are 124 samples total: 16 who had a family history of breast cancer and carried a BRCA mutation and had developed breast cancer, 23 with a family history of breast cancer but did not carry a BRCA1/2 mutation and had developed breast cancer, 22 with a family history of breast cancer and carried a BRCA1/2 mutation but had not developed breast cancer, 22 with a family history of breast cancer but did not carry a BRCA1/2 mutation and had not developed breast cancer, 22 with no family history of breast cancer and had developed breast cancer, and 19 with no family history of breast cancer and had not developed breast cancer. The Ontario cohort included 73 samples: 11 who had a family history of breast cancer and carried a BRCA mutation and had developed breast cancer, 17 with a family history of breast cancer but did not carry a BRCA1/2 mutation and had developed breast cancer, 14 with a family history of breast cancer and carried a BRCA1/2 mutation but had not developed breast cancer, 18 with a family history of breast cancer but did not carry a BRCA1/2 mutation and had not developed breast cancer, 8 with no family history of breast cancer and had developed breast cancer, and 5 with no family history of breast cancer and had not developed breast cancer. 36 of the Ontario samples were scanned and processed at Duke University; the remaining 37 samples were handled at Boston University. The batch-adjustment process was repeated twice independently: [1] all 124 Utah samples were adjusted jointly with the 36 Ontario samples that had been processed at Duke University (set 1) and [2] all 124 Utah samples were adjusted jointly with the 37 Ontario samples that had been processed at Boston University (set 2). This dataset represents set 1.