Hi, I've been applying FIRMA on the Gardina et.al (http:// www.biomedcentral.com/1471-2164/7/325) Human Exon Array dataset which has been summarized based on mappings from the BrainArray CustomCDFs (in particular ENSE13). Taking the list of genes which have been validated through PCR, I found the corresponding probesets on the customCDF which interrogate the differentially spliced exons. When I plot an ROC curve for the positive and negative probesets according to the paper using the FIRMA scores, the ROC curve shows poor results by FIRMA (auc =0.4949367)
When I do the analysis using the official CDF from affymetrix sticking to the positive and negatives probesets listed in the FIRMA paper by Purdom et. al (http://bioinformatics.oxfordjournals.org/content/ 24/15/1707.abstract), the ROC curve does very well (auc=0.89) I find this puzzling considering that the customCDF from the Brainarray group is considered to be more accurate since they filter out all the problematic probesets (eg. multi-mapping). One would expects the FIRMA scores to be better when using these re-annotated CDF. I am thinking it might be because the probesets might contain less probes in them than before and thus when fitting RMA and calculating the residuals from the fitting, the residuals might be skewed or less accurate due to lesser probes contributing to the FIRMA score? Does anyone have any ideas as to why FIRMA does so poorly with the customCDF? Thanks, Fong -- When reporting problems on aroma.affymetrix, make sure 1) to run the latest version of the package, 2) to report the output of sessionInfo() and traceback(), and 3) to post a complete code example. You received this message because you are subscribed to the Google Groups "aroma.affymetrix" group with website http://www.aroma-project.org/. To post to this group, send email to aroma-affymetrix@googlegroups.com To unsubscribe and other options, go to http://www.aroma-project.org/forum/
