Good morning, Sorry for the trouble. I am trying to apply a pamr gene classifier obtained from Affymetrix data onto spot cDNA array data. For the Affy data, we have tumors , as well as several normals. The spot cDNA data is tumor against normal. I was hoping that I could get the small gene classifier, somehow (black box) modify this classifier with the mean value of the corresponding gene from normal tissue, and apply this with pamr.predict onto the spot data. I don't really know what Gaussian linear discrimination is, so I am a little lost in the original paper, but it seems that I should be modifying the discriminant scores somehow to construct estimates of the class probabilities. I tried running pamr with all Affymetrix data converted to ratios. Results are definitely not as good as just using the rma expression values, which is my preferred option. I would appreciate any insights on this problem. Thank you! Min-Han This email message, including any attachments, is for the so...{{dropped}}

On Tue, Sep 28, 2004 at 01:04:00AM -0400, Tan, MinHan wrote: > Good morning, > > Sorry for the trouble. > > I am trying to apply a pamr gene classifier obtained from Affymetrix > data onto spot cDNA array data. For the Affy data, we have tumors , as > well as several normals. The spot cDNA data is tumor against normal. > > I was hoping that I could get the small gene classifier, somehow (black > box) modify this classifier with the mean value of the corresponding > gene from normal tissue, and apply this with pamr.predict onto the spot > data. > > I don't really know what Gaussian linear discrimination is, so I am a > little lost in the original paper, but it seems that I should be > modifying the discriminant scores somehow to construct estimates of the > class probabilities. Hi, In general it doesn't make too much sense to develop a model based on one set of measurements (one type of measurements) on a different set of measurements. I would be quite skeptical of any results that did not involve some fairly rigorous description of why (and under what conditions) such a comparison would be sensible. I believe that the first step should be deciding what comparisons between your two data sets (cDNA and Affy) are sensible, then you might try to find a model that captures those aspects of the data. You might look at papers like Parmigiani et al, Clinical Cancer Research, 2004, or Choi et al, Bioinformatics, 2003. Both consider issues in combining or comparing microarray studies. Robert > > I tried running pamr with all Affymetrix data converted to ratios. > Results are definitely not as good as just using the rma expression > values, which is my preferred option. > > I would appreciate any insights on this problem. Thank you! > > Min-Han > > This email message, including any attachments, is for the so...{{dropped}} > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor -- +--------------------------------------------------------------------- ------+ | Robert Gentleman phone : (617) 632-5250 | | Associate Professor fax: (617) 632-2444 | | Department of Biostatistics office: M1B20 | | Harvard School of Public Health email: rgentlem@jimmy.harvard.edu | +--------------------------------------------------------------------- ------+