Figure 1

MA and cumulative distribution function (CDF) plots. MA-plots are log expression in treatmentminus (M) log expression in control versus average (A) log expression plots. (a) For two sets of triplicates from the Affymetrix HGU133A spike-in experiment [2,3] we calculated the average log ratio across the three comparisons (M) and the average log intensity (A) across all six arrays for each feature. The figure shows M plotted against A. However, because there are hundreds of thousands of features, instead of plotting each point, we use shades of blue to denote the amount of points in each region of the plot. About 90% of the data is contained in the dark-blue regions. Orange points are the 405 features from the 36 genes with nominal fold changes of 2. (b) As in (a) but using two sets of biological triplicates from a study comparing three trisomic human brains to three normal human brains. The orange dots are 385 features representing 35 genes on chromosome 21 for which we expect fold changes of 1.5. (c) Empirical cumulative density functions for the median corrected log (base 2) intensities of 50 randomly chosen arrays from the Gene Expression Omnibus (GEO), three randomly selected arrays from Affymetrix HGU133A spike-in experiment, and the three S samples from Choe et al. [1] facilitate the comparison; the intensities were made to have the same median. The dashed black horizontal lines show the 75% and 95% percentiles. (d) As (a) but showing the two sets of triplicates described by Choe et al. [1]. The orange dots are 375 features randomly sampled from those that were spiked-in to have fold changes greater than 1. The yellow ellipse is used to illustrate an artifact: among the data with nominal fold changes of 1, there appear to be two clusters having different overall observed log ratios.