Figure 2

The effect of different preprocessing methods. Smoothed scatter plots showing data from a typical sample before (A) and after (B and C) different preprocessing methods (B: within-channel quantile normalization and C: quantile normalization (as in B) combined with loess correction to remove the intensity dependent curvature evident in panels A and B). In these plots, darker regions indicate a higher density of points. Panels D to F show beanplots [20] of the log-ratios from 20 samples for the different methods. Between sample consistency improves after normalization (E and F). Data shown is from the Omni1-Quad platform.