Figure 3

Functional gene clusters are significantly deregulated. (A) Distributions of the deregulation scores of the genes in the functional clusters (averaged over the three regulators, ATM, RelA and p53) strongly deviate from the distribution of averaged deregulation scores for all genes (left plot). The distributions of average decorrelation scores (middle left), as well as of average regulation scores in the healthy (middle right) and in the damaged cells (right), are more similar to the distributions of the same scores for all genes. Gray dashed vertical lines mark score 0 in each plot. The names of the eleven damage-activated clusters are in bold. (B) A t-test comparing the cluster deregulation scores with the deregulation scores for all genes (Deregulation; red squares) gives for majority of the clusters the most significant p-values, when contrasted with: the p-values obtained in a t-test comparing cluster regulation scores to regulation scores of all genes in the healthy cells (Regulation, Healthy; blue pluses), and the same t-test but in the damaged cells (Regulation, Damage; light blue crosses), the p-values in a t-test comparing cluster regulation scores in the healthy directly to regulation scores in the damaged cells (Healthy vs Damage; green diamonds), and in a t-test comparing the cluster decorrelation scores with the decorrelation scores for all genes (Decorrelation; yellow triangles). All tests are two-sided. Taken together, our joint and knowledge-based approach assigns more significant scores to the functional clusters than a separate analysis, or an analysis without incorporation of knowledge.