Figure 5 in the recent Nucleic Acids Research 2003, Vol. 31, No. 4 e15 (http://nar.oupjournals.org/cgi/content/full/31/4/e15/GNG015F5) shows an interesting asymmetry in the distribution of fold changes for non- spiked in genes. Can anyone comment on the reason for this? Here, dChip has the least amount of bias. Thanks, Chris

I don't think that this is really a bias. There are probesets that have been observed to perform much like the "official" spikeins for various reasons. See http://affycomp.biostat.jhsph.edu/whymore.html for details. These probesets would be giving the asymmetry out in the extremes of the plot you referred to. Ben On Mon, 2003-08-11 at 14:47, Chris Paulse wrote: > Figure 5 in the recent Nucleic Acids Research 2003, Vol. 31, No. 4 e15 > (http://nar.oupjournals.org/cgi/content/full/31/4/e15/GNG015F5) shows an > interesting asymmetry in the distribution of fold changes for non- spiked in > genes. Can anyone comment on the reason for this? Here, dChip has the > least amount of bias. > > Thanks, > Chris > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor -- Ben Bolstad <bolstad@stat.berkeley.edu> http://www.stat.berkeley.edu/~bolstad

remember this only relates to the range of the 12000+ genes. most of the data is in the orange/yellow area. as described in this list and here: http://affycomp.biostat.jhsph.edu/ after that paper was published, some have found 2 other genes that appear to be spike-ins that affymetrix forgot to report. if you redefine those as true positives the bias you describe goes away and the range goes down significantly for both. because pm-only dchip attenuates signal more than rma, the bias appears bigger for RMA if you include the new found spikeins as false positives. becuase of MAS 5's huge variance at the low end the range is not affected by considering these two genes as false positives. plots similar to Figure 5 in the NAR paper can be found in the affycomp web page stated above (see Figure 6a and 6b). hope this helps, rafael On Mon, 11 Aug 2003, Chris Paulse wrote: > Figure 5 in the recent Nucleic Acids Research 2003, Vol. 31, No. 4 e15 > (http://nar.oupjournals.org/cgi/content/full/31/4/e15/GNG015F5) shows an > interesting asymmetry in the distribution of fold changes for non- spiked in > genes. Can anyone comment on the reason for this? Here, dChip has the > least amount of bias. > > Thanks, > Chris > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >