Fig. 1

Profile expression deconvolution. a Background normalized transcripts according to P, M, and A flag values streamlined Affymetrix probeset populations that were further refined to report transcripts with significant expression changes (P <0.01). The resulting transcript dynamics were profiled to examine magnitude and direction of expression changes, which revealed diverse gene expression patterns. b Kohonen mapping, or Self-Organizing Maps (SOM), categorized the transcriptome into 12 discrete clusters, visualized as a U-matrix. Multidimensional data visualized by U-matrices are organized into hexagon plots of alternating nodes (containing gene cluster metaprofiles) and non-nodes (indicating similarity). Each node is numbered and presents the metagene expression profile for that cluster, while non-nodes indicate percent similarity between clusters according to Euclidean distance metrics. Grayscale bar indicates high (white) to low (black) percent similarity. c Top pathways prioritized for each cluster, with p-value indicated for each