Genome-wide mRNA studies in breast cancer have revealed the heterogeneity of breast cancer and helped to characterize the distinct disease states. The gene expression subtypes of breast cancer are associated with prognosis and response to therapy. Thus, subtype diagnosis could influence care, but there exists no clinically available test for subtype assignment. Therefore, the goal of this research was to develop such an assay for biological discovery, and to asses its clinical utility. Clinical classification requires a measurement assay and accompanying quantitative model whose performance has been tested with relevant material using contemporary patient cohorts. A 50 gene qRT-PCR assay was developed using clinically available formalin-fixed paraffin-embedded materials, thus allowing utilization of existing archives. An objective subtype assignment algorithm was simultaneously developed and assessed in a large cohort of patients who received no adjuvant systemic therapy. The assay and algorithm demonstrated prognostic value in old age tissue providing proof of principal for further investigation. The first blinded evaluation used an ER-positive, tamoxifen treated cohort of archived specimens with up to 15 years of follow-up information. Results of this study confirm strong association between subtype assignment and clinical outcomes. These associations validate the available information content, and confirm that the test is capable of identifying subjects for whom no further therapy is needed. This identification is not currently possible, thus the assay demonstrates clear clinical utility in a contemporary cohort. The above studies demonstrate that gene expression-based assignments provide information regarding prognosis and therapeutic response. DNA copy number aberrations (CNA) are also a likely data source of clinical and biological value, and thus, this additional information content was evaluated. Results illustrate specific tumor DNA copy number aberrations hold prognostic value. While these aberrations do not improve on gene expression based models, the CNA model explains similar prognostic variation in luminal tumors and thus may aid by identifying candidate drug targets in these tumors. Based on these studies we have developed a clinical assay for subtype diagnosis, demonstrated that the assay is more informative than clinical markers, and provide the first step toward elucidating tumorigenic pathways in luminal tumors.