Searches for over- or under-represented genes for GO categories.
This tool takes a filtered data set and goes through all GO terms, no matter on what level of the GO hierarchy they are. Analysis can be restricted to GO terms which have a minimum, user-specified number of genes mapped to them (by default 5). In order to avoid identifying directly related GO terms with a considerable overlap of genes, the tool performs conditional testing by default. This conditional method first tests those GO terms, which don't have any child terms. If these terms are significant at a specified p- value cutoff, the genes that are annotated with this term are removed before testing the parent terms. This continues until all terms have been tested. It is also possible to run the tool using unconditional testing. When the tool is run that way, multiple testing correction can be applied.
Output is a text file and an HTML file with a list of significant categories. If you would like to know what are the genes listed as "observed", you can run the tool "Utilities/ Extract genes from GO term" on your original data set. Please note that this can sometimes yield a higher number of rows than the gene number indicated in the hypergeomteric test result. This is because there are several probes/probesets for one gene, and they are treated as one gene in the hypergeometric test.