Evaluation of Microarray Preprocessing Algorithms Based on Concordance with RT-PCR in Clinical Samples
Figure 1
The colon and cell line data sets are representative of clinical microarray data.
For several Affymetrix data sets, box-and-whiskers plots indicate the distribution of three bias metrics: a) RNA degradation slope, b) median perfect-match probe intensity, and c) fraction of probe sets called present. A narrower distribution indicates greater consistency in technical conditions. LatinSquare133 and LatinSquare95 are spike-in data sets produced by the microarray manufacturer [1]; Gyorffy_cells and Gyorffy_colon are the data sets analyzed in this paper [11], [12]; the other five are publicly-available clinical data sets [14]–[18].