Hi Parantu The boxplot looks for me. Because of the higher variation in the low expression range, VST transformation shrinks the values in this region, which is also reflected in the boxplot. A following up normalization is required after VST transformation. As for which normalization method to use, it also depends on the data set and the purpose of the study. Hope this is helpful to you. Pan On 9/21/09 9:13 AM, "parantu shah" <parantu.shah at="" gmail.com=""> wrote: > Hi. > > I am using Lumi package for analyzing Illumina gene expression array > data. I don't have control data and I am using sample probe > profile.txt file. > > ?I use 'force positive' and 'bgAdjust.affy' options to do background > correction. After doing VST and RSN normalization - estimate of differentially > expressed genes (using limma) ?is very different (1675 vs. 1134). > > I am attaching you the box plot of VST data after no correction, > forcepositive and bgAdjust.affy in a hope that it will be useful to > you in answering. > > I would also appreciate any other suggestions you have. > > Thanks and best regards, > Parantu. > > > -- > Parantu Shah, PhD > Dept. of Biostatistics & Computational Biology > Dana-Farber Cancer Institute > Harvard School of Public Health > > CLS-11075, ?3 Blackfan Circle > Boston MA 02115 > Phone : 617 582 8852 > http://www.hsph.harvard.edu/~pshah > >

Hi. I am using Lumi package for analyzing Illumina gene expression array data. I don't have control data and I am using sample probe profile.txt file. ?I use 'force positive' and 'bgAdjust.affy' options to do background correction. After doing VST and RSN normalization - estimate of differentially expressed genes (using limma) ?is very different (1675 vs. 1134). I am attaching you the box plot of VST data after no correction, forcepositive and bgAdjust.affy in a hope that it will be useful to you in answering. I would also appreciate any other suggestions you have. Thanks and best regards, Parantu. -- Parantu Shah, PhD Dept. of Biostatistics & Computational Biology Dana-Farber Cancer Institute Harvard School of Public Health CLS-11075, ?3 Blackfan Circle Boston MA 02115 Phone : 617 582 8852 http://www.hsph.harvard.edu/~pshah -- Parantu Shah, PhD Dept. of Biostatistics & Computational Biology Dana-Farber Cancer Institute Harvard School of Public Health CLS-11075, 3 Blackfan Circle Boston MA 02115 Phone : 617 582 8852 http://www.hsph.harvard.edu/~pshah -------------- next part -------------- A non-text attachment was scrubbed... Name: bgcorr.pdf Type: application/pdf Size: 43203 bytes Desc: not available URL: <https: stat.ethz.ch="" pipermail="" bioconductor="" attachments="" 20090921="" 15dd56c0="" attachment.pdf="">

Hi Parantu, As Dr. Du said, the background correction, transformation, and normalization steps have a complex impact on the end-results. There are a few publications (NAR) on the topic of background correction, but I think further research is still necessary. If you are interested in a follow up investigation, we would certainly be happy to facilitate. Best regards, Simon *************************************************** Simon Lin, MD, CSDP Director of Bioinformatics Consulting 750 N. Lake Shore Drive, 11th Floor, Room 174 Research Associate Professor Northwestern University Biomedical Informatics Center/NUCATS Chicago, IL 60611 S-Lin2 at northwestern.edu Phone 312-503-2353 Fax 312-503-5388 > Hi Parantu > > The boxplot looks for me. Because of the higher variation in the low > expression range, VST transformation shrinks the values in this region, > which is also reflected in the boxplot. A following up normalization is > required after VST transformation. As for which normalization method to use, > it also depends on the data set and the purpose of the study. > Hope this is helpful to you. > > > Pan > > > On 9/21/09 9:13 AM, "parantu shah" <parantu.shah at="" gmail.com=""> wrote: > >> Hi. >> >> I am using Lumi package for analyzing Illumina gene expression array >> data. I don't have control data and I am using sample probe >> profile.txt file. >> >> I use 'force positive' and 'bgAdjust.affy' options to do background >> correction. After doing VST and RSN normalization - estimate of differentially >> expressed genes (using limma) is very different (1675 vs. 1134). >> >> I am attaching you the box plot of VST data after no correction, >> forcepositive and bgAdjust.affy in a hope that it will be useful to >> you in answering. >> >> I would also appreciate any other suggestions you have. >> >> Thanks and best regards, >> Parantu. >> >> >> -- >> Parantu Shah, PhD >> Dept. of Biostatistics & Computational Biology >> Dana-Farber Cancer Institute >> Harvard School of Public Health >> >> CLS-11075, 3 Blackfan Circle >> Boston MA 02115 >> Phone : 617 582 8852 >> http://www.hsph.harvard.edu/~pshah >> >> > > >