Gene expression profiles distinguishing good from poor responders. (a) Hierarchical clustering of 37 genes that defined the class predictor and whose expression best differentiated good from poor responders before chemotherapy. Each row represents a single gene and each column represents the average of two available independent FNAs. Green and red squares indicate, respectively, overexpressed and underexpressed genes in a breast tumor compared with the MCF10A breast cancer cell line (color intensity is proportional to the magnitude of the expression level ratio). Black squares indicate genes with approximately equivalent expression levels and gray squares indicate missing or filter-excluded data. Branches representing good responders are shown in blue and those representing poor responders in yellow. (b) Hierarchical clustering of 16 genes whose change in expression best differentiated good from poor responders after one cycle of chemotherapy.