Hello list, I am using the package "topGO" to analyse GO enrichment of gene sets: My genes are ensembl IDs and are not taken from a microarray, so I had to feed "topGOdata" with a gene2GO list. (see http://thread.gmane.org/gmane.science.biology.informatics.conductor/14 627) I construct that list by mapping all ensembl IDs to GO IDs using the package "biomaRt". Then I proceed with my analysis: > GOdata <- new("topGOdata", ontology = "MF", allGenes = selectedList, description = "Ensembl GO enrichment", annot = annFUN.gene2GO, gene2GO = gene2GO) Do you confirm this approach is correct? I also had several question concerning topGO: - Are the p-value in topGO corrected for multiple testing (FDR...)? My guess is that they are not due to a problem of independence... - Are there some differences between Fisher exact test (topGO) and Hypergeometric test (GOstats). If yes, why did the two packages make different choices? - It is not clear to me what the Kolmogorov-Smirnov is testing? Especially in my case where I don't provide scores associated to my genes... - Is there a way to test separately over/under representation of GO categories? Thanks a lot in advance for your help or tips Julien -- Julien Roux, PhD student http://www.unil.ch/dee/page22707.html Department of Ecology and Evolution Biophore, University of Lausanne, 1015 Lausanne, Switzerland tel: +41 21 692 4221 fax: +41 21 692 4165

Hi Julien, Julien Roux wrote: > Hello list, > > I am using the package "topGO" to analyse GO enrichment of gene sets: > > My genes are ensembl IDs and are not taken from a microarray, so I had > to feed "topGOdata" with a gene2GO list. > (see > http://thread.gmane.org/gmane.science.biology.informatics.conductor/ 14627) > I construct that list by mapping all ensembl IDs to GO IDs using the > package "biomaRt". > Then I proceed with my analysis: > > > GOdata <- new("topGOdata", ontology = "MF", allGenes = selectedList, > description = "Ensembl GO enrichment", annot = annFUN.gene2GO, gene2GO = > gene2GO) > > Do you confirm this approach is correct? It should be correct. You simply need a named character vector where the names are the Entrez Gene IDs, and the vector contains GO IDs. > > I also had several question concerning topGO: > - Are the p-value in topGO corrected for multiple testing (FDR...)? My > guess is that they are not due to a problem of independence... I don't think they are corrected. I'm not even sure you could (or should). As with a lot of microarray analyses, p-values should not be taken at face value. Rather they should be used more as ranking tools. > - Are there some differences between Fisher exact test (topGO) and > Hypergeometric test (GOstats). If yes, why did the two packages make > different choices? Both packages are using the same test. The Fisher exact test is used to assess association between variables in a 2x2 contingency table. Under the null hypothesis of independence the counts in a given table follow a hypergeometric distribution, so the p-values for a 2x2 table are computed using this distribution. See e.g., ?fisher.test > - It is not clear to me what the Kolmogorov-Smirnov is testing? > Especially in my case where I don't provide scores associated to my genes... > - Is there a way to test separately over/under representation of GO > categories? In GOstats there is. I don't know about topGO. Best, Jim > > Thanks a lot in advance for your help or tips > Julien > -- James W. MacDonald, M.S. Biostatistician Affymetrix and cDNA Microarray Core University of Michigan Cancer Center 1500 E. Medical Center Drive 7410 CCGC Ann Arbor MI 48109 734-647-5623