Fig. 2

Gene expression analysis of five cancer cell lines with five total RNA concentrations. A Responses to increasing amounts of total RNA from five different cancer cell lines for probes to five different genes. In the absence of cell line RNA (no template control samples with 0 pg of total RNA), there were only very small signals on the probes. For each probe the medians of nine no template control samples were subtracted from the cell line data to obtain net signals. The gene symbols and the cell lines are indicated at the top of each plot. The assay was performed in triplicate using separate aliquots of total RNA. Individual replicates are plotted in black, and the mean values of the three replicates are plotted in green. Regression lines were fit to the mean values, and the slopes of the regression lines are indicated on the plots. The correlation coefficients (R) and the p-values of the correlation coefficients are also indicated on the plots. Note that in some cases, 20 pg of RNA produced very large signals (nearing saturation), indicating high expression of the gene within certain cell lines. B Principal Components Analysis of the data from 53 genes assayed with five different amounts of total RNA from five different cell lines. The data were mean centered but not scaled, so some genes may have had more influence on clustering than others; clustering of the cell lines was somewhat better without scaling. C Hierarchical clustering of the data from 53 genes assayed with 20 pg of total RNA from five different cell lines. Data were log2 transformed, and then the data for each probe in the columns were mean centered and scaled by dividing by the standard deviation for each probe. Colors represent relative Z-Scores. The distance method used for hierarchical clustering was non-parametric rank order Kendall correlation, and the linkage method used was Ward.D2