Fig. 1

Global analysis of differential expression. (a) GM6 structure. (b) Hierarchical cluster analysis. FC estimates for 6777 protein-coding genes are shown in the yellow-blue heat map (see scale) and by lines on the right (red = 6 h; green = 24 h; blue = 48 h). Genes were clustered using average linkage and the Euclidean distance metric. The 6777 genes represent 50% of 13,554 genes included in all differential expression analyses (selected as having the largest absolute FC estimate in any of the 4 differential expression analyses). (c) Principal component (PC) vectors (2-dimensional). Arrows represent the effect of GM6. Arrow start points indicate the average PC coordinates of CTL samples and arrow end points indicate the average PC coordinates of GM6 samples. (d) PC vectors (3-dimensional). The analysis in (c) was repeated using the first 3 PC axes. (e) FC correlation between time points. Scatterplots compare FC estimates for the 3 possible pairwise time point comparisons. The color scale indicates gene density. The spearman rank correlation is indicated in the lower right for each analysis. (f) Self-organizing map (SOM) color images. A SOM was calculated based upon FC estimates observed for 12,562 protein-coding genes. The color scale indicates the average FC for genes assigned to each SOM region. (g) Smoothed SOM surfaces. The SOMs from (f) are alternatively represented by smoothed surfaces with the vertical axis corresponding to the average FC of genes assigned to each SOM region. Rotations are shown for each surface to provide alternative viewing angles (45, 90, 135, 180 and 225 degrees). The smoothed surface was obtained by fitting a loess function using average FC as the response variable and the other two axes as predictor variables. (h – j) Percent GM6-increased genes (FC > 1.00) per chromosome. Asterisks indicate chromosomes with a significantly large (red) or small (blue) percentage of GM6-increased genes (FDR < 0.05; Fisher’s Exact Test)