This study aimed to classify breast carcinomas based on gene expression patterns from cDNA microarrays and correlate tumor characteristics with clinical outcomes. Using hierarchical clustering, 85 cDNA microarray experiments representing 78 cancers, three fibroadenomas, and four normal breast tissues were analyzed. The cancers were classified into three main groups: basal epithelial-like, ERBB2-overexpressing, and normal breast-like. A novel finding was that the previously characterized luminal epithelial/estrogen receptor-positive group could be divided into at least two subgroups with distinct expression profiles. These subtypes showed robust clustering using different gene sets and were associated with different clinical outcomes. The study found that tumors classified based on gene expression patterns could serve as prognostic markers for overall and relapse-free survival in a subset of patients. The luminal epithelial/estrogen receptor-positive group was further divided into two subgroups, luminal subtype A and B/C, with distinct gene expression profiles and different prognoses. The basal-like subtype was associated with poor prognosis, while the ERBB2+ subtype showed significant differences in outcome. TP53 mutations were more frequent in the ERBB2+ and basal-like subtypes, suggesting a role for TP53 in determining gene expression patterns. The study also identified that tumors classified as luminal A had better outcomes than luminal B + C, with luminal B + C tumors showing a worse disease course and higher relapse rates. The findings highlight the importance of gene expression patterns in understanding breast cancer biology and clinical outcomes. The study provides evidence that five expression-based subclasses of breast tumors are associated with patient outcomes, with particular interest in the finding that ER+ tumors may be subclassified into distinct subgroups with different outcomes. These results suggest that gene expression patterns can be used to identify clinically relevant subtypes of breast cancer, which may lead to improved treatment strategies.This study aimed to classify breast carcinomas based on gene expression patterns from cDNA microarrays and correlate tumor characteristics with clinical outcomes. Using hierarchical clustering, 85 cDNA microarray experiments representing 78 cancers, three fibroadenomas, and four normal breast tissues were analyzed. The cancers were classified into three main groups: basal epithelial-like, ERBB2-overexpressing, and normal breast-like. A novel finding was that the previously characterized luminal epithelial/estrogen receptor-positive group could be divided into at least two subgroups with distinct expression profiles. These subtypes showed robust clustering using different gene sets and were associated with different clinical outcomes. The study found that tumors classified based on gene expression patterns could serve as prognostic markers for overall and relapse-free survival in a subset of patients. The luminal epithelial/estrogen receptor-positive group was further divided into two subgroups, luminal subtype A and B/C, with distinct gene expression profiles and different prognoses. The basal-like subtype was associated with poor prognosis, while the ERBB2+ subtype showed significant differences in outcome. TP53 mutations were more frequent in the ERBB2+ and basal-like subtypes, suggesting a role for TP53 in determining gene expression patterns. The study also identified that tumors classified as luminal A had better outcomes than luminal B + C, with luminal B + C tumors showing a worse disease course and higher relapse rates. The findings highlight the importance of gene expression patterns in understanding breast cancer biology and clinical outcomes. The study provides evidence that five expression-based subclasses of breast tumors are associated with patient outcomes, with particular interest in the finding that ER+ tumors may be subclassified into distinct subgroups with different outcomes. These results suggest that gene expression patterns can be used to identify clinically relevant subtypes of breast cancer, which may lead to improved treatment strategies.