...

To review, including a factor in the middle box will identify GO categories whose expression is consistently affected across the genes within the category by the factor of interest. Including a factor in the right box (factor*gene) will identify gene ontology categories where the expression of the genes within the category are affected but not uniformly across the genes within the category.

Contrasts

GO ANOVA is not restricted to analysis of factors with only two levels. The ANOVA p-values are measures of likelihood that all groups are equivalent. While this is useful in general, sometimes tests comparing only two sets of data are more desirable. Using contrasts to define pair wise comparisons in an ANOVA model is superior to using a test that is limited to a two group comparison.

To specify individual pair wise comparisons, press the Contrast button. Contrasts are performed on groups already defined in the ANOVA model. If two tissue types should be compared to each other, select the tissue term from the Select Factor/Interaction dropdown in the upper left. Select either one or a set of categories and add them to group 1 and group 2. All samples falling into group 1 will be compared to all samples falling into group 2. Output will include not only a p-value, but also a fold change. This fold change will represent the average fold change of the GO category between the two groups. Fold change is calculated as Group 1 divided by Group 2. For data in log space, the data is antilogged as well; fold change output is always for data on a linear scale.