Figure 3.

Comparison of A/P calls using ROC curves. Different models of probe-specific background are used as the basis for generating A/P calls of gene expression. We plot the ROC curve as the true positive rate versus the false positive rate of agreement between each A/P call method and gold-standard sets of expressed and unexpressed genes generated from independent SAGE data. The ROC curves from several A/P methods are shown here in each of three tissues, (a) cerebellum, (b) heart, (c) liver: Exon array MAT background (red); Exon array Affymetrix DABG (blue); maximum 3' array MAS 5.0 probeset statistic (purple); minimum 3' array MAS5.0 probeset statistic (brown).