Effect of the amount of noise in SAGE data on the performance of seriation and PoissonC.

Seriation and PoissonC were applied to a simulated SAGE data set containing three expression patterns and increasing amount of noise tags. The dataset is described in more detail in the text and in Table S1. TP (True Positives) include tags that were correctly classified as belonging to the correct expression group (expression pattern 1, 2, or 3 or noise) by assigning them to the cluster (PoissonC) or contig (seriation) containing other members of the expression group. FP (False Positives) include noise tags that have been erroneously assigned to a cluster or contig with tags that conform to the expression pattern 1, 2, or 3.
*The false positive rate is significantly higher for the PoissonC algorithm than it is for seriation mostly due to the erroneous assignment of noise tags to an expression pattern (p<0.05).