Patterns of gene expression which predict clinical outcome in breast cancer have been identified, but there is clinical heterogeneity in breast cancer. Integration of clinical parameters with expression profiling could identify more homogenous subgroups. Dai et al. identified a subset of breast cancer patients, those with relatively high estrogen receptor expression for their age, who had a fairly homogeneous gene expression pattern as a function of clinical outcome. Expression profiling of snap frozen breast cancer specimens from patients in this subgroup was analyzed by an iterative training sample selection method to reveal the dominant pattern of gene expression associated with metastasis. In the cover image of a the patients (rows) with metastases (solid circles) cluster at the bottom of the figure and tend to have overexpression (magenta) of a common set of genes (columns). This gene set consists almost entirely of cell cycle genes. Over-expression of this set of genes is much less correlated with clinical outcome in other patient subpopulations. These results suggest that cell proliferation is a driving force to extremely poor outcome in one subpopulation of patients and that combining clinical and molecular parameters could be of value in refining other gene expression predictors of clinical outcome. For details, see the article by Dai et al. on page 4059 of this issue.