Cell lines derived from human tumors have historically served as the primary experimental model system for exploration of tumor cell biology and pharmacology. Cell line studies, however, must be interpreted in the context of artifacts introduced by selection and establishment of cell lines {\it in vitro}. This complication has led to difficulty in the extrapolation of biology observed in cell lines to tumor biology {\it in vivo}. Modern genomic analysis tool like DNA microarrays and gene expression profiling now provide a platform for the systematic characterization and classification of both cell lines and tumor samples. Studies using clinical samples have begun to identify classes of tumors that appear both biologically and clinically unique as inferred from their distinctive patterns of expressed genes. In this review, we explore the relationships between patterns of gene expression in breast tumor derived cell lines to those from clinical tumor specimens. This analysis demonstrates that cell lines and tumor samples have distinctive gene expression patterns in common and underscores the need for careful assessment of the appropriateness of any given cell line as a model for a given tumor subtype.