Illustration of the microarrays results. Raw data files from the feature extraction software for image analysis were imported into the Resolver system for gene expression data analysis from Rosetta Informatics. Then combined experiments were generated to obtain average values from the replicates and dye-swap experiments in order to avoid dye incorporation bias. Microarray data were processed and combined using the Rosetta Resolver system, as described.¹⁹  (A) Hierarchical cluster analysis. (1) 416 probes: expression profiles in the 4N, 8N, and 16N MK populations compared to the 2N MK population. (2) 943 probes: expression profiles in the 2N, 4N, 8N, and 16N MK populations compared with an MK pool. ANOVA was performed on the 41 059 probes of the chips with a P threshold of < .001. These genes were subjected to hierarchical cluster analysis using a calculation based on Pearson correlation and agglomerative method on the average link. Each row represents the combination of 2 dye-swap experimental samples, and each column represents a single accession number. Three hundred sixty-seven probes are common between these 2 analyses. (B) Analysis of the discriminating genes between the 2N-4N and 8N-16N ploidy levels. Combined microarray data were exported from Resolver into BRB ArrayTools. Analysis of discriminating genes between the 2N-4N and the 8N-16N groups was performed using the class prediction module with a random variance model. As data were previously normalized by feature extraction software, no spot filtering and normalization were performed. Genes with greater than 50% missing values were excluded. Finally, 105 genes were selected based on a P threshold of < .001. (red squares) Overexpression versus the pool of all samples. (green squares) Underexpression versus the pool of all samples. Gene names are displayed as RefSeq or GenBank identifiers on the left, whereas they are displayed as gene symbols (or probe names) on the right. (C) Illustration of the 3 clusters of genes with a ploidy-regulated expression. Three clusters from 8 representing the genes for which the expression is down- or up- regulated with ploidy. Up-regulated genes are divided into 2 clusters, one with a linear increase of expression level between different classes of ploidy (cluster I) and another reaching a plateau between 8N and 16N (cluster II). The y axis represents the LogRatio between the intensity of hybridization of each ploidy level (2N, 4N, 8N, or 16N) and the intensity of hybridization of the MK pool. The x-axis represents the different ploidy levels.