On Fri, Apr 01, 2005 at 08:04:28PM -0500, James MacDonald wrote: > It doesn't make any sense to use gcrma() if you don't have MM probes; > the idea behind gcrma is to come up with a better measure of background > than the MM measure itself. A modification of gcrma() that doesn't use > MM probes is rma(). And if you are using a tiling array it does not seem to make sense (to me at least) to use rma, since tiling arrays does not have the cpncept of probesets. But I do not know your particular array, so I may be wrong. Kasper > >>> Shinhan Shiu <shiu@uchicago.edu> 04/01/05 5:13 PM >>> > We are trying to use GCRMA to adjust raw intensity values from tiling > chip > experiments (Arabidopsis). But the affy Arabidopsis tiling chip do not > have > mismatch probes and it seems the mismatch probe intensity is absolutely > required in: > > bg.parameters.ns > > Where the mismatch probe intensities, mismatch probe affinity, and > perfect > match probe affinities are passed. I wonder how this function can be > modified so only perfect match probe info is used. Thanks. > > Shinhan > > > ******************************** > Shinhan Shiu > Dept. of Ecology and Evolution > University of Chicago > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > ********************************************************** > Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues. > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor -- Kasper Daniel Hansen, Research Assistant Department of Biostatistics, University of Copenhagen

To estimate non-specific binding, some negative control probes are desired. However, arrays do not have to have the MM probes as in Affymetrix GeneChip design (one for each PM probe) as negative control probes. They simply should be probes that do not match any specific target in your sample. The function bg.parameters.ns estimates the relationship between probe affinities and non-specific binding(NSB) with one set of negative control probes (in Affy chips, the MMs), and predicts NSB in another set of probes (in Affy chips, the PMs). So as long as you have some negative controls, you can modify the parameters passed to "bg.parameters.ns" accordingly. If you have no negative control probes at all, you probably need some extra assumption, otherwise the NSB and SB may not be identified. If you can assume a large proportion of specific target to be absent, you can also simply pass PM affinities and PM intensities instead of the MMs to bg.parameters.ns(). > At 12:43 PM 4/4/2005, you wrote: > >Hi Shinhan; > >In GCRMA, bg.adjust.affinities use PM and MM intentisities to fit a model > >to extract the signal ( what we call background correction). This step > >used sequence information for affinity calculation. Without MM > >intensities, the model is not complete or there is no model at all, > >meaning this step is meaningless. So if I understand correctly, it is > >impossible to adjust affinities for PM probes without MM information. > > > >For details, reference paper: "A model based background adjustment for > >oligonucleotide expression arrays" by Zhijin Wu etc. > > > >Fangxin > > > > > > > > > I should have been more specific. What I want to do is to use the affinity > > > calculated by gcrma to account for GC contents of the probes. So I am > > > not planning to use the rma part of GCRMA. > > > > > > gcrma first calls compute.affinities to generate affinity for each probe. > > > After optical correction (bg.adjust.optical), gcrma calls gcrma.engine > > > which calls bg.adjust.affinities. This is the method I am interested in. I > > > was hoping to use this method to correct the raw intensity for probe GC > > > contents. > > > > > > But I found that bg.adjust.affinities calls bg.parameters.ns that requires > > > MM probe intensity, MM probe affinity, and PM probe affinities. I am a bit > > > surprised because, as James commented, I thought gcrma don't need MM > > > information. > > > > > > What I would like to find out is: how I should pass the parameter or > > > modify > > > the method for adjusting affinities for PM probes without MM information. > > > > > > Shinhan > > > > > > At 01:09 AM 4/2/2005, Kasper Daniel Hansen wrote: > > >>On Fri, Apr 01, 2005 at 08:04:28PM -0500, James MacDonald wrote: > > >> > It doesn't make any sense to use gcrma() if you don't have MM probes; > > >> > the idea behind gcrma is to come up with a better measure of > > >> background > > >> > than the MM measure itself. A modification of gcrma() that doesn't use > > >> > MM probes is rma(). > > >> > > >>And if you are using a tiling array it does not seem to make sense (to > > >>me at least) to use rma, since tiling arrays does not have the cpncept > > >>of probesets. > > >> > > >>But I do not know your particular array, so I may be wrong. > > >> > > >>Kasper > > >> > > >> > >>> Shinhan Shiu <shiu@uchicago.edu> 04/01/05 5:13 PM >>> > > >> > We are trying to use GCRMA to adjust raw intensity values from tiling > > >> > chip > > >> > experiments (Arabidopsis). But the affy Arabidopsis tiling chip do not > > >> > have > > >> > mismatch probes and it seems the mismatch probe intensity is > > >> absolutely > > >> > required in: > > >> > > > >> > bg.parameters.ns > > >> > > > >> > Where the mismatch probe intensities, mismatch probe affinity, and > > >> > perfect > > >> > match probe affinities are passed. I wonder how this function can be > > >> > modified so only perfect match probe info is used. Thanks. > > >> > > > >> > Shinhan > > >> > > > >> > > > >> > ******************************** > > >> > Shinhan Shiu > > >> > Dept. of Ecology and Evolution > > >> > University of Chicago > > >> > > > >> > _______________________________________________ > > >> > Bioconductor mailing list > > >> > Bioconductor@stat.math.ethz.ch > > >> > https://stat.ethz.ch/mailman/listinfo/bioconductor > > >> > > > >> > > > >> > > > >> > ********************************************************** > > >> > Electronic Mail is not secure, may not be read every day, and should > > >> not be used for urgent or sensitive issues. > > >> > > > >> > _______________________________________________ > > >> > Bioconductor mailing list > > >> > Bioconductor@stat.math.ethz.ch > > >> > https://stat.ethz.ch/mailman/listinfo/bioconductor > > >> > > >>-- > > >>Kasper Daniel Hansen, Research Assistant > > >>Department of Biostatistics, University of Copenhagen > > > > > > ******************************** > > > Shinhan Shiu > > > Dept. of Ecology and Evolution > > > University of Chicago > > > > > > _______________________________________________ > > > Bioconductor mailing list > > > Bioconductor@stat.math.ethz.ch > > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > > > > > > > > >-- > >Fangxin Hong, Ph.D. > >Plant Biology Laboratory > >The Salk Institute > >10010 N. Torrey Pines Rd. > >La Jolla, CA 92037 > >E-mail: fhong@salk.edu > > ******************************** > Shinhan Shiu > Dept. of Ecology and Evolution > University of Chicago > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >

Hi, Shinhan, I am afraid I do not have a number to give you. Background will be over estimated when you pass PM intensity and PM affinity instead of MM, as you would expect. When you assume that majority of the PMs mainly contain NSB, hopefully the functional relationship between log(PM-o)~affinity resembles that of NSB~affinity. In fact the MM probes detect specific binding(SB) as well but to a much lower extent, compared to to PMs. So it seems that the SB in MM probes did not affect the estimates of NSB much. How much SB is there in your sample is hard to estimate if you have absolutely no negative control probes. You may get some idea by hybridizing one chip to a non-specific sample with which you expect no signal. best, Jean On Mon, 4 Apr 2005, Shinhan Shiu wrote: > Hi Zhijin, > > Thanks for the message. I have a question on the second solution: passing > only the PM intensity and PM affinity information to bg.parameter.ns. You > mentioned that the assumption in excluding MM information is when "a large > proportion of specific targets are absent". I wonder if you can clarify > this a little. Do you mean that "the intensities of a large proportion of > the probes are simply background"? Also, do you have some estimate as to > how large a proportion is sufficient? > > Also, does this approach (excluding MM info) lead to a higher false > negative rate when comparing to using negative controls for MM? Thanks again. > > Shinhan > > At 01:18 PM 4/4/2005, you wrote: > >To estimate non-specific binding, some negative control probes are > >desired. However, arrays do not have to have the MM probes as in > >Affymetrix GeneChip design (one for each PM probe) as negative control > >probes. They simply should be probes that do not match any specific target > >in your sample. > > The function bg.parameters.ns estimates the relationship between probe > >affinities and non-specific binding(NSB) with one set of negative control > >probes (in Affy chips, the MMs), and predicts NSB in another set of probes > >(in Affy chips, the PMs). So as long as you have some negative controls, > >you can modify the parameters passed to "bg.parameters.ns" accordingly. If > >you have no negative control probes at all, you probably need some > >extra assumption, otherwise the NSB and SB may not be identified. If you > >can assume a large proportion of specific target to be absent, you can > >also simply pass PM affinities and PM intensities instead of the MMs to > >bg.parameters.ns(). > > > > > > > At 12:43 PM 4/4/2005, you wrote: > > > >Hi Shinhan; > > > >In GCRMA, bg.adjust.affinities use PM and MM intentisities to fit a model > > > >to extract the signal ( what we call background correction). This step > > > >used sequence information for affinity calculation. Without MM > > > >intensities, the model is not complete or there is no model at all, > > > >meaning this step is meaningless. So if I understand correctly, it is > > > >impossible to adjust affinities for PM probes without MM information. > > > > > > > >For details, reference paper: "A model based background adjustment for > > > >oligonucleotide expression arrays" by Zhijin Wu etc. > > > > > > > >Fangxin > > > > > > > > > > > > > > > > > I should have been more specific. What I want to do is to use the > > affinity > > > > > calculated by gcrma to account for GC contents of the probes. So I am > > > > > not planning to use the rma part of GCRMA. > > > > > > > > > > gcrma first calls compute.affinities to generate affinity for each > > probe. > > > > > After optical correction (bg.adjust.optical), gcrma calls gcrma.engine > > > > > which calls bg.adjust.affinities. This is the method I am > > interested in. I > > > > > was hoping to use this method to correct the raw intensity for probe GC > > > > > contents. > > > > > > > > > > But I found that bg.adjust.affinities calls bg.parameters.ns that > > requires > > > > > MM probe intensity, MM probe affinity, and PM probe affinities. I > > am a bit > > > > > surprised because, as James commented, I thought gcrma don't need MM > > > > > information. > > > > > > > > > > What I would like to find out is: how I should pass the parameter or > > > > > modify > > > > > the method for adjusting affinities for PM probes without MM > > information. > > > > > > > > > > Shinhan > > > > > > > > > > At 01:09 AM 4/2/2005, Kasper Daniel Hansen wrote: > > > > >>On Fri, Apr 01, 2005 at 08:04:28PM -0500, James MacDonald wrote: > > > > >> > It doesn't make any sense to use gcrma() if you don't have MM > > probes; > > > > >> > the idea behind gcrma is to come up with a better measure of > > > > >> background > > > > >> > than the MM measure itself. A modification of gcrma() that > > doesn't use > > > > >> > MM probes is rma(). > > > > >> > > > > >>And if you are using a tiling array it does not seem to make sense (to > > > > >>me at least) to use rma, since tiling arrays does not have the cpncept > > > > >>of probesets. > > > > >> > > > > >>But I do not know your particular array, so I may be wrong. > > > > >> > > > > >>Kasper > > > > >> > > > > >> > >>> Shinhan Shiu <shiu@uchicago.edu> 04/01/05 5:13 PM >>> > > > > >> > We are trying to use GCRMA to adjust raw intensity values from > > tiling > > > > >> > chip > > > > >> > experiments (Arabidopsis). But the affy Arabidopsis tiling chip > > do not > > > > >> > have > > > > >> > mismatch probes and it seems the mismatch probe intensity is > > > > >> absolutely > > > > >> > required in: > > > > >> > > > > > >> > bg.parameters.ns > > > > >> > > > > > >> > Where the mismatch probe intensities, mismatch probe affinity, and > > > > >> > perfect > > > > >> > match probe affinities are passed. I wonder how this function can be > > > > >> > modified so only perfect match probe info is used. Thanks. > > > > >> > > > > > >> > Shinhan > > > > >> > > > > > >> > > > > > >> > ******************************** > > > > >> > Shinhan Shiu > > > > >> > Dept. of Ecology and Evolution > > > > >> > University of Chicago > > > > >> > > > > > >> > _______________________________________________ > > > > >> > Bioconductor mailing list > > > > >> > Bioconductor@stat.math.ethz.ch > > > > >> > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > >> > > > > > >> > > > > > >> > > > > > >> > ********************************************************** > > > > >> > Electronic Mail is not secure, may not be read every day, and should > > > > >> not be used for urgent or sensitive issues. > > > > >> > > > > > >> > _______________________________________________ > > > > >> > Bioconductor mailing list > > > > >> > Bioconductor@stat.math.ethz.ch > > > > >> > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > >> > > > > >>-- > > > > >>Kasper Daniel Hansen, Research Assistant > > > > >>Department of Biostatistics, University of Copenhagen > > > > > > > > > > ******************************** > > > > > Shinhan Shiu > > > > > Dept. of Ecology and Evolution > > > > > University of Chicago > > > > > > > > > > _______________________________________________ > > > > > Bioconductor mailing list > > > > > Bioconductor@stat.math.ethz.ch > > > > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > > > > > > > > > > > > > > > > > > >-- > > > >Fangxin Hong, Ph.D. > > > >Plant Biology Laboratory > > > >The Salk Institute > > > >10010 N. Torrey Pines Rd. > > > >La Jolla, CA 92037 > > > >E-mail: fhong@salk.edu > > > > > > ******************************** > > > Shinhan Shiu > > > Dept. of Ecology and Evolution > > > University of Chicago > > > > > > _______________________________________________ > > > Bioconductor mailing list > > > Bioconductor@stat.math.ethz.ch > > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > > > > ******************************** > Shinhan Shiu > Dept. of Ecology and Evolution > University of Chicago > ******************************** > >