Four binary classifications of high grade ovarian serous cancer.

<p>The ISIS algorithm identified four independent binary partition classifications (splits) of 129 ovarian cancer samples. Each binary classification is supported by an independently selected set of 100 genes (module). The top panel of this figure shows four horizontal bars representing the classification of the 129 tumor samples (columns) with respect to the gene modules. Red indicates that a patient was classified into the smaller group resulting from the respective split (g1) and white indicates the classification into the larger group (g0). The heatmap in the lower panel represents the expression profiles of the gene modules supporting the four binary classifications. Each row represents a gene, each column a patient and each cell correspond to a gene and its expression level; yellow indicates an expression level of a gene above its mean across the patients and blue below its mean.