Cross Study Analysis

Cross Study Gene Analysis
A cross study 'consensus' fold change was calculated for each gene and disease/demographic comparison. The consensus fold change is based on a weighted combination of the individual fold changes and standard error for the probes that map to each gene across the platforms/studies. The weights are equal to 1/SE_i, where SE_i is the standard error of the i'th probe for the gene. For disease fold changes, 95% confidence intervals are calculated, and for demographic factors the 99% CI is used.

Cross Study Pathway/GO Analysis
As described here, Pathway and GO relative risks were calculated for each pathway/GO term and each study/disease combination. Corresponding z-scores were also computed as a measure of significance. The study centric pathway/GO disease plots combine the data for all of these analyses into a set of heatmap displays, where cells are colored by z-score. Cross study z-scores were also computed from the individual study results.

In contrast, the gene centric pathway/GO disease plots first combine the data at the gene level to form consensus fold changes, as described above. Consensus fold changes that are significant at the .05 level are assessed by forming the two-by-two contingency tables for each pathway/GO term (regulated/unregulated vs. in pathway/not in pathway) , then testing for significance of that association (by Fisher's exact test). The results are then displayed in heatmaps colored by p-value.

The demographic pathway/GO plots are generated similarly to the gene-centric disease plots, the only differnce being that the definition of significance for individual genes was p<.01.