Microarray Normalization Revisited for Reproducible Breast Cancer Biomarkers

Figure 1

Comparison of pipelines. Each pipeline is given a token (e.g., RmaMGlobal) for reference throughout the paper. Specific pairs of pipelines have been selected to be compared in detail regarding the differences in output they produce. Each of these comparisons is labelled by a capital letter (“A” to “N”) for later reference. Note that only a reasonable selection (and not all possible) comparisons have been considered in detail. Moreover, all possible comparisons are quantified by the underlayed heat map, giving the percent misclassification in prediction of oestrogene receptor status, see the colorbar on the right. This heat map shows the values given in the first line (ER) of Table 2.