Figure 2

Quantification of dye effects. Smoothed scatter plots of the estimated dye effect from the SNP-wise linear models for the non-normalized gDNA (A) and cDNA (B) data versus average intensity are shown. In these plots, a higher density of points is represented by a darker shade of blue. For each analysis, the log-ratios are calculated as log₂(Cy 5/Cy 3). Probes at lower intensities tend to have a bias towards the Cy3 channel (negative dye effect), while probes with higher intensities generally have a bias towards the Cy5 channel (positive dye-bias). Panel C shows the estimated dye effect before and after quantile normalization. The dye effect is systematically larger, and more variable before quantile normalization for both sample types. After quantile normalization, the dye effects are centered around zero.