Figure 2. Pairwise and principal component analysis of microarray data.

(A–B) Scatter-plots of the normalized probe intensity of all annotated microarray signals are shown. Each dot represents the mean of 6 independent biological replicates. Numbers in the box represent the number of features that were at least 2-fold different at a 99% confidence interval by moderated t test with Benjamini-Hochberg false discovery rate analysis. Data are plotted at a log₂ scale. Panel (A) compares NOD.Rag^−/− mice versus NOD mice at 2 wks of age and panel (B) compares them at 6 wks of age. (C) Hierarchically clustered heat map (Euclidean row distance) of the 24 differentially expressed transcripts identified in (B). (D–E) Scatter-plots as in (A–B) except the comparison is between 2 and 6 wk old NOD or 2 and 6 wk NOD.Rag^−/−. (F) GO terms and transcription factor binding analysis of the differences identified in (D–E). The two graphs on the left, labeled ‘Shared’ show signatures common to both NOD and NOD.Rag^−/−. The signatures on the right, labeled ‘NOD only’, show changes specific to the NOD strain. (G) Principal component analysis of microarray samples. Each of 57 microarrays is summarized as a point and drop-line. Samples are color coded as indicated by the title of each group. (H) Hierarchical clustered heat map (Euclidean distance) of the top 1% most variant genes identified by principal component. (i) GO and transcription binding analysis of the top 1% most variant genes identified by principal component analysis.