This figure summarizes a very large set of measurements in which the mRNA levels of 1800 selected genes (arranged top to bottom) were determined for 142 different human tumors (arranged left to right), each from a different patient. Each small red bar indicates that the given gene in the given tumor is transcribed at a level significantly higher than the average across all the cell lines. Each small green bar indicates a less-than-average expression level, and each black bar denotes an expression level that is close to average across the different tumors. The procedure used to generate these data—mRNA isolation followed by hybridization to DNA microarrays—is described in Chapter 8 (see pp. 533–535). The figure shows that the relative expression levels of each of the 1800 genes analyzed vary among the different tumors (seen by following a given gene left to right across the figure). This analysis also shows that each type of tumor has a characteristic gene expression pattern. This information can be used to “type” cancer cells of unknown tissue origin by matching the gene expression profiles to those of known tumors. For example, the unknown sample in the figure has been identified as a lung cancer. (Courtesy of Patrick O. Brown, David Botstein, and the Stanford Expression Collaboration.)