Meta Analysis

From Array Suite Wiki

Array Server provides the user with the ability to run a meta analysis on a number of different projects, either matching row name (for exact matches in the same platform), or by matching by Master ID (as well as an option to match by region for testing across differing platforms (i.e. SNP vs. expression). This is useful for testing a number of related projects, and looking for genes that may be significant when taking all of these projects into account.


The user must choose 2 or more inference reports upon which to run the meta analysis. The primary data (which will be the first inference report chosen) will be used for matching purposes. Clicking the "Add" button opens the Auto-fill window, where you can quickly filter projects by meta data. In the example below, "neoplasm" was used to match Project ID's:


Selecting a Project ID will display associated Inference report test data. Selecting "OK" will add the project to the Meta Analysis window. Repeat this process to add additional projects.


The user has a number of options for running the analysis:

  • First, the analysis can be mapped using Row ID (for direct platform comparisons), master ID (for comparisons across platforms within a data type (i.e. snp vs. snp or expression vs. expression).
  • The user can set the Score cutoff. This uses the same scoring system as Search Variable Profile. The idea is, results are returned if a variable has an average score over the cutoff. Therefore, variables will be returned that have high scores (top p-values relative to that project) in both projects. The Score cutoff can be anything over 90.

The results are a report tab, showing each combination of primary data/secondary data, for a particular variable, that had a score above the cutoff. In this case, there were only a few probesets that had high scores.

Clicking the Expand all button in the Task tab will expand all the results.


Take a look at the first example. The score for this result is 99.68, because that is the average of the highest contrast score for each of the compared projects (score of 99.84 first project blue symbol and a score of 94.49 for the contrast in the second project green symbol).


Clicking the links will take you to a VariableData view, or show information about the annotation for that gene (masterID). In addition, a table is returned showing a flat structure for the meta analysis (this can be used for exporting).