Breast tumor subtypes correlate with prognosis

To advance in our understanding of the biological pathways involved in breast cancer tumor progression we have analyzed a set of breast tumor biopsies in order to identify the genomic pathways in which tumor may develop. With this objective, a cDNA microarray platform containing 800 genes was constructed. These genes were chosen because they are in several representatives signaling pathways, namely estrogen and progesterone receptor related pathways, cell cycle, DNA repair, chromatin remodeling, cell proliferation, apoptosis, cell adhesion, cell invasion and angiogenesis. An analysis of the gene expression profile of a large group of breast tumors was carried out and good correlation with their clinical-histopathological data was obtained. We identified different tumor subclusters with distinctive phenotype characteristics within our population, which later correlated with clinical outcome. The most significant genes able to discriminate between different tumor phenotypes of our training set of samples were determined applying a “Leave-one-out” cross-validation method of statistical analysis called PAM (Prediction analysis of microarrays). The predictor was further tested on a new incoming set of samples where we determined to which subtype of tumor new samples were allocated and predicted outcome was compared to clinical data, showing how some distinct tumor phenotypes correlate with a poor prognosis. Pathway analysis of the most significant genes belonging to each phenotype was performed to elucidate the most representative biological signaling pathways through which tumor progression might elapse. We constructed a real time expression profiling platform with a selection of the most differentially expressed genes in each subtype, which could be used as a technique to help improve the diagnosis and prognosis of the breast tumor samples of our sample population. A set of 105 breast tumor tissue samples and 6 from normal breast tissue adjacent to tumors of different individuals were amplified, labeled and hybridized, as described in the methods to our cDNA platform. Universal human reference RNA (UHRR) was used as reference RNA. Breast tumor samples were labeled with Cy5 and the universal human reference RNA was labeled with Cy3. There are no dye-swaps.