Hello, I'm new in the analysis of ChIP-chip data, and have some questions. My data: ChIP-chip dataset on post-translational modification of the histone protein H4 (acetylation on K12) in human sperm. Microarray platform: NimbleGen promoter design, 385K two-array set. I use the Ringo package to analyse the raw data. My questions: - it is possible to find ChIP-enriched regions without smoothing the reporter intensities? - which statistical analysis are adequate after finding ChIP-enriched regions? - there is any algorithm in Bioconductor similar/comparable to the one use by Nimblegen for peaks finding (FDR calculation)? I will very appreciate some advices. Thanks on advance Dr. Viviana Menzel Germany -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Viviana Menzel Rottweg 34 35428 Langg?ns Tel.: +49 6403 7748550 Mobil: +49 177 5126092 E-Mail: vivianamenzel at gmx.de Web: www.dres-menzel.de

Hello, please find a few answers below your questions. On Thu, 21 Jan 2010 11:58:27 +0100, Dr. Viviana Menzel wrote > Hello, > > I'm new in the analysis of ChIP-chip data, and have some questions. > My data: ChIP-chip dataset on post-translational modification of the > histone protein H4 (acetylation on K12) in human sperm. Microarray > platform: NimbleGen promoter design, 385K two-array set. > I use the Ringo package to analyse the raw data. My questions: > - it is possible to find ChIP-enriched regions without smoothing the > reporter intensities? The aim of the smoothing step is to down-weigh the individual reporter effects which add noise to the reporter measurements of actual ChIP enrichment in the respective genomic regions. I would recommend to perform this step to get more accurate measurements of enrichment and estimates for the null distribution of reporter intensities under non-enrichment. > - which statistical analysis are adequate after finding ChIP- enriched > regions? That really depends on what you want to do with the enriched regions. For example, you can relate the enriched regions to genes and then check whether the list of affected genes is significantly associated to a biological theme, as described in the GO or KEGG databases. The packages topGO or GOstats would be useful for such an analysis. > - there is any algorithm in Bioconductor similar/comparable to the > one use by Nimblegen for peaks finding (FDR calculation)? Besides the peak-finding methods in Ringo, there are also peak-finding methods in other packages, for example in ACME and BAC. One caveat may be that, with histone modification, you would normally not expect the triangular peak shape that is characteristic for transcription-factor peaks, so some of the methods might be less useful for your data. You can also have a look at this article in which I have described the analysis of ChIP-chip with Ringo in a bit more detail than in the vignette. http://www.ploscollections.org/article/info%3Adoi%2F10.1371%2Fjournal. pcbi.1000227 The data package ccTutorial accompanies this article. I am sure that other people from the list can provide you with additional ideas how best to analyse your data. Best regards, Joern --- Joern Toedling Institut Curie -- U900 26 rue d'Ulm, 75005 Paris, FRANCE Tel. +33 (0)156246927