Figure 5

Stability of determined log-ratio against signal intensity. The upper part of the panel is an example of a rank vs. log-ratio plot for data comparison [5]. Each channel of the data was normalized by means of the three-parameter lognormal distribution model prior to the comparison. In the plot, the x-axis shows the ranks of average z-scores, and the y-axis shows the log-ratio that was calculated by means of the difference in z-scores. Intensity dependence of the determined log-ratio can be found by tilting and/or varying the width of the bands formed by the dots (see Results for details), as well as by noting changes in the moving averages (green line) and moving standard deviations (blue line). Red colored dots indicate that one of the paired data is in the model-inconsistent range. In the lower part of the panel, distributions of measurements are presented in histograms at the indicated ranges of ranks. Variations in intensity dependence can be visualized in the particular shape, height, width, and/or the center value of such histograms. Data were obtained from 0 hr and 24 hr after serum shock treatment [21]; the total number of data was 9677, and the experiment ID was 5731 [20]. In the control and treatment experiments, data ranked below -1.4 and above 2.2 were considered inconsistent with the distribution model (red dots). The 300 lowest-intensity data were excluded as "not detected" since they were found to be negative through γ subtraction. The window of the moving average and standard deviation calculation was 200.