Figure 7From: Estimation and correction of non-specific binding in a large-scale spike-in experimentPlot of mean log2 difference versus mean log2 intensity (MA plot) between C and S samples for probesets normalized after loess and vsn. MA plots are for probesets with GC-NSB correction, vsn probe-level normalization, medianpolish probe summary and (a) loess or (b) vsn probeset-level normalization. The vsn probeset-level normalization method is better than the loess method at separating TPs from FPs (AUC score of 0.916 for loess and 0.928 for vsn), but the null FC = 1 probesets after vsn probeset-level normalization are clearly not centered around zero and the method has more than 25% more FPs with q-values below 0.10. Only probesets with present transcripts are shown (FC = 1, black; FC > 1, red).Back to article page