Taqman array analysis

0

Entering edit mode

james perkins ▴ 300

@james-perkins-2675

Last seen 10.3 years ago

Hi, Apologies for the long list of questions, I have searched the mailing list but can't find much info about these arrays. I am looking at low density PCR cards. They measure the expression levels of 96 different transcripts from a very small sample of human or animal tissue. There are actually 384 reactions going on but in our case each is done in quadruplicate (can be through biological or technical repetition). I wondered if there was a favoured way to normalise this data. The most cited paper I have found is the Vandesompele 2002 paper using the geometric mean of a number of control genes, implemented in R in the SLqPCR. Has anything else been developed that could be used with these cards? I guess quantile normalisation is out of the question since this makes some assumption that the majority of genes don't change in expression. In addition, does anything exist in bioconductor (or outside it) to identify and remove outlying data points? The cards work by having a series of microfluidic channels deliver samples to 384 well PCR reactions. Sometimes an air bubble or something else means that the odd reaction fails. Also is there a favoured way to determine what is consistently different between control and experimental samples. I assume a False Discovery Rate method is still in favour, possibly from t-test (or LIMMA??) but we are also interested in fold-change. Currently I just mean each gene and divide case by control to get a crude measure of fold change. Kind regards, James Perkins

• 2.2k views

ADD COMMENT • link updated 16.3 years ago by Bas Jansen ▴ 150 • written 16.3 years ago by james perkins ▴ 300

0

Entering edit mode

Bas Jansen ▴ 150

@bas-jansen-2966

Last seen 10.3 years ago

Hi James: On Mon, Sep 1, 2008 at 1:25 PM, James Perkins <jperkins at="" biochem.ucl.ac.uk=""> wrote: > Hi, > > > Apologies for the long list of questions, I have searched the mailing list > but can't find much info about these arrays. > > > I am looking at low density PCR cards. They measure the expression levels of > 96 different transcripts from a very small sample of human or animal tissue. > There are actually 384 reactions going on but in our case each is done in > quadruplicate (can be through biological or technical repetition). > > I wondered if there was a favoured way to normalise this data. The most > cited paper I have found is the Vandesompele 2002 paper using the geometric > mean of a number of control genes, implemented in R in the SLqPCR. > > Has anything else been developed that could be used with these cards? I > guess quantile normalisation is out of the question since this makes some > assumption that the majority of genes don't change in expression. As far as I know nothing has been developed in Bioconductor for these cards. When I analyzed them, I first created an ExpressionSet following the (excellent!) directions given in the the Biobase vignette 'An introduction to Bioconductor's ExpressionSet class' by Falcon et al. Then I processed the normalized data (deltaCt) using the LMGene package in order to perform gene-by-gene ANOVA and to identify differentially expressed genes. I have repeated the whole procedure using different control genes (read: different deltaCt values for the same gene), but in my case I got the same results with the different controls. Hope this helps. Kind regards, Bas

ADD COMMENT • link 16.3 years ago Bas Jansen ▴ 150

0

Entering edit mode

Hi Bas, Thanks for your reply. I have built an eset with detector as the rows and sample as the columns. However I have not been able to populate it with delta Ct since I do not have this data. How did you calculate deltaCt? Using the proprietary software? I don't have access to this I have just been given the Ct and the Ct Avg for each detector. I have been normalising each gene to the houskeeping genes, averaging across samples and dividing case by control to get the fold change. I've then been comparing the resultant fold changes depending on choice of normaliser against each other to see if there is a difference, which there is with *some* control genes. Kind regards, James Bas Jansen wrote: > Hi James: > > On Mon, Sep 1, 2008 at 1:25 PM, James Perkins > <jperkins at="" biochem.ucl.ac.uk=""> wrote: > >> Hi, >> >> >> Apologies for the long list of questions, I have searched the mailing list >> but can't find much info about these arrays. >> >> >> I am looking at low density PCR cards. They measure the expression levels of >> 96 different transcripts from a very small sample of human or animal tissue. >> There are actually 384 reactions going on but in our case each is done in >> quadruplicate (can be through biological or technical repetition). >> >> I wondered if there was a favoured way to normalise this data. The most >> cited paper I have found is the Vandesompele 2002 paper using the geometric >> mean of a number of control genes, implemented in R in the SLqPCR. >> >> Has anything else been developed that could be used with these cards? I >> guess quantile normalisation is out of the question since this makes some >> assumption that the majority of genes don't change in expression. >> > > As far as I know nothing has been developed in Bioconductor for these cards. > When I analyzed them, I first created an ExpressionSet following the > (excellent!) directions given in the the Biobase vignette 'An > introduction to Bioconductor's ExpressionSet class' by Falcon et al. > Then I processed the normalized data (deltaCt) using the LMGene > package in order to perform gene-by-gene ANOVA and to identify > differentially expressed genes. I have repeated the whole procedure > using different control genes (read: different deltaCt values for the > same gene), but in my case I got the same results with the different > controls. Hope this helps. > > Kind regards, > Bas >

ADD REPLY • link 16.3 years ago james perkins ▴ 300

0

Entering edit mode

On Thu, Sep 4, 2008 at 6:05 AM, James Perkins <jperkins at="" biochem.ucl.ac.uk=""> wrote: > Hi Bas, > > Thanks for your reply. I have built an eset with detector as the rows and > sample as the columns. However I have not been able to populate it with > delta Ct since I do not have this data. > > How did you calculate deltaCt? Using the proprietary software? I don't have > access to this I have just been given the Ct and the Ct Avg for each > detector. There are a number of references for this (quick google search will turn up several). But here is one that is pretty clear, I think. Livak, K. J. and Schmittgen, T. D. (2001). Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402-8. http://www.ncbi.nlm.nih.gov/pubmed/11846609 Sean > Bas Jansen wrote: >> >> Hi James: >> >> On Mon, Sep 1, 2008 at 1:25 PM, James Perkins >> <jperkins at="" biochem.ucl.ac.uk=""> wrote: >> >>> >>> Hi, >>> >>> >>> Apologies for the long list of questions, I have searched the mailing >>> list >>> but can't find much info about these arrays. >>> >>> >>> I am looking at low density PCR cards. They measure the expression levels >>> of >>> 96 different transcripts from a very small sample of human or animal >>> tissue. >>> There are actually 384 reactions going on but in our case each is done in >>> quadruplicate (can be through biological or technical repetition). >>> >>> I wondered if there was a favoured way to normalise this data. The most >>> cited paper I have found is the Vandesompele 2002 paper using the >>> geometric >>> mean of a number of control genes, implemented in R in the SLqPCR. >>> >>> Has anything else been developed that could be used with these cards? I >>> guess quantile normalisation is out of the question since this makes some >>> assumption that the majority of genes don't change in expression. >>> >> >> As far as I know nothing has been developed in Bioconductor for these >> cards. >> When I analyzed them, I first created an ExpressionSet following the >> (excellent!) directions given in the the Biobase vignette 'An >> introduction to Bioconductor's ExpressionSet class' by Falcon et al. >> Then I processed the normalized data (deltaCt) using the LMGene >> package in order to perform gene-by-gene ANOVA and to identify >> differentially expressed genes. I have repeated the whole procedure >> using different control genes (read: different deltaCt values for the >> same gene), but in my case I got the same results with the different >> controls. Hope this helps. >> >> Kind regards, >> Bas >> > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor >

ADD REPLY • link 16.3 years ago Sean Davis 21k

0

Entering edit mode

Hi James, by normalising to housekeeper genes, you have probably inadvertently calculated deltaCt. There are some other great references by Pfaffl and Bustin on the subject cheers, Mark ----------------------------------------------------- Mark Cowley, BSc (Bioinformatics)(Hons) Peter Wills Bioinformatics Centre Garvan Institute of Medical Research, Sydney, Australia ----------------------------------------------------- On 04/09/2008, at 8:05 PM, James Perkins wrote: > Hi Bas, > > Thanks for your reply. I have built an eset with detector as the > rows and sample as the columns. However I have not been able to > populate it with delta Ct since I do not have this data. > > How did you calculate deltaCt? Using the proprietary software? I > don't have access to this I have just been given the Ct and the Ct > Avg for each detector. > > I have been normalising each gene to the houskeeping genes, > averaging across samples and dividing case by control to get the > fold change. I've then been comparing the resultant fold changes > depending on choice of normaliser against each other to see if there > is a difference, which there is with *some* control genes. > > Kind regards, > > James > > Bas Jansen wrote: >> Hi James: >> >> On Mon, Sep 1, 2008 at 1:25 PM, James Perkins >> <jperkins at="" biochem.ucl.ac.uk=""> wrote: >> >>> Hi, >>> >>> >>> Apologies for the long list of questions, I have searched the >>> mailing list >>> but can't find much info about these arrays. >>> >>> >>> I am looking at low density PCR cards. They measure the expression >>> levels of >>> 96 different transcripts from a very small sample of human or >>> animal tissue. >>> There are actually 384 reactions going on but in our case each is >>> done in >>> quadruplicate (can be through biological or technical repetition). >>> >>> I wondered if there was a favoured way to normalise this data. The >>> most >>> cited paper I have found is the Vandesompele 2002 paper using the >>> geometric >>> mean of a number of control genes, implemented in R in the SLqPCR. >>> >>> Has anything else been developed that could be used with these >>> cards? I >>> guess quantile normalisation is out of the question since this >>> makes some >>> assumption that the majority of genes don't change in expression. >>> >> >> As far as I know nothing has been developed in Bioconductor for >> these cards. >> When I analyzed them, I first created an ExpressionSet following the >> (excellent!) directions given in the the Biobase vignette 'An >> introduction to Bioconductor's ExpressionSet class' by Falcon et al. >> Then I processed the normalized data (deltaCt) using the LMGene >> package in order to perform gene-by-gene ANOVA and to identify >> differentially expressed genes. I have repeated the whole procedure >> using different control genes (read: different deltaCt values for the >> same gene), but in my case I got the same results with the different >> controls. Hope this helps. >> >> Kind regards, >> Bas >> > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

ADD REPLY • link 16.3 years ago Mark Cowley ▴ 910

0

Entering edit mode

Thanks everyone for your replies. I was indeed inadvertently calculating deltaCt, but its great to know that what I was doing was correct! I get differences in values for deltaCt using different control genes, but the trend is always the same and these changes are usually quite small, except for a few occasions where there seems to have been problems determining the value for Ct in some of the technical replicates. (so if GAPDH had a Ct value of ~24 cycles for 2 technical replicates, but was undetermined for 2 others). However I guess this is more of a problem with the technology and possibly the person running the card than with the choice of genes. Regards, Jim Mark Cowley wrote: > Hi James, > by normalising to housekeeper genes, you have probably inadvertently > calculated deltaCt. > There are some other great references by Pfaffl and Bustin on the subject > > cheers, > Mark > > ----------------------------------------------------- > Mark Cowley, BSc (Bioinformatics)(Hons) > > Peter Wills Bioinformatics Centre > Garvan Institute of Medical Research, Sydney, Australia > ----------------------------------------------------- > > On 04/09/2008, at 8:05 PM, James Perkins wrote: > >> Hi Bas, >> >> Thanks for your reply. I have built an eset with detector as the rows >> and sample as the columns. However I have not been able to populate >> it with delta Ct since I do not have this data. >> >> How did you calculate deltaCt? Using the proprietary software? I >> don't have access to this I have just been given the Ct and the Ct >> Avg for each detector. >> >> I have been normalising each gene to the houskeeping genes, averaging >> across samples and dividing case by control to get the fold change. >> I've then been comparing the resultant fold changes depending on >> choice of normaliser against each other to see if there is a >> difference, which there is with *some* control genes. >> >> Kind regards, >> >> James >> >> Bas Jansen wrote: >>> Hi James: >>> >>> On Mon, Sep 1, 2008 at 1:25 PM, James Perkins >>> <jperkins at="" biochem.ucl.ac.uk=""> wrote: >>> >>>> Hi, >>>> >>>> >>>> Apologies for the long list of questions, I have searched the >>>> mailing list >>>> but can't find much info about these arrays. >>>> >>>> >>>> I am looking at low density PCR cards. They measure the expression >>>> levels of >>>> 96 different transcripts from a very small sample of human or >>>> animal tissue. >>>> There are actually 384 reactions going on but in our case each is >>>> done in >>>> quadruplicate (can be through biological or technical repetition). >>>> >>>> I wondered if there was a favoured way to normalise this data. The >>>> most >>>> cited paper I have found is the Vandesompele 2002 paper using the >>>> geometric >>>> mean of a number of control genes, implemented in R in the SLqPCR. >>>> >>>> Has anything else been developed that could be used with these >>>> cards? I >>>> guess quantile normalisation is out of the question since this >>>> makes some >>>> assumption that the majority of genes don't change in expression. >>>> >>> >>> As far as I know nothing has been developed in Bioconductor for >>> these cards. >>> When I analyzed them, I first created an ExpressionSet following the >>> (excellent!) directions given in the the Biobase vignette 'An >>> introduction to Bioconductor's ExpressionSet class' by Falcon et al. >>> Then I processed the normalized data (deltaCt) using the LMGene >>> package in order to perform gene-by-gene ANOVA and to identify >>> differentially expressed genes. I have repeated the whole procedure >>> using different control genes (read: different deltaCt values for the >>> same gene), but in my case I got the same results with the different >>> controls. Hope this helps. >>> >>> Kind regards, >>> Bas >>> >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor at stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >

ADD REPLY • link 16.3 years ago james perkins ▴ 300

Login before adding your answer.