Begin forwarded message: > From: Sean Davis <sdavis2@mail.nih.gov> > Date: October 1, 2004 1:29:57 PM EDT > To: "Matthew Hannah" <hannah@mpimp-golm.mpg.de> > Subject: Re: [BioC] ULYSIS Alexa Fluor labelling kit / dye bias / limma > > Matt, > > Ah, I see what you are saying. Yes, limma can (and so can many other > non-center-based normalization methods like loess) remove the > dependence of ratio on amplitude rather nicely. However, this DOES > NOT remove dye bias from any given point. For example, if one has a > spot that has an intensity in the green channel of 200 and 400 in the > red channel (for a ratio of 2) and one does the dye swap, the > expectation is that the ratio will be 0.5, but for dye-biased spots, > the ratio may be 1 (or something other than 0.5). If this is > repeatable, then the spot is said to have dye bias--the expression of > the spot is biased by dye. One can only discover this by looking at > at least two arrays, and they must, of course, be in dye swap. > Normalization does not account for this effect. > > Sean > > On Oct 1, 2004, at 1:05 PM, Matthew Hannah wrote: > >> Sean, >> From the limma normalisewithinarrays help >> >> This function normalizes M-values (log-ratios) for dye-bias within >> each >> array. Apart from method="none" and method="median", all the >> normalization methods make use of the relationship between dye-bias >> and >> intensity. The loess normalization methods were proposed by Yang et al >> (2001, 2002). Smyth and Speed (2003) give a detailed statement of the >> methods. >> >> But I have no idea how this would effect things if there was less/no >> bias - hence the question. >> >> Sorry I remembered reading it but should have mentioned this in the >> post >> rather than being abit vague. >> >> Cheers, >> Matt >> >>> -----Original Message----- >>> From: Sean Davis [mailto:sdavis2@mail.nih.gov] >>> Sent: Freitag, 1. Oktober 2004 18:48 >>> To: Matthew Hannah >>> Subject: Re: [BioC] ULYSIS Alexa Fluor labelling kit / dye >>> bias / limma >>> >>> Matt, >>> >>> I don't think limma directly accounts for dye bias in most >>> (all?) of its normalization methods. You can't remove dye >>> bias with a within-slide normalization. You can remove dye >>> bias from your analysis using a linear model, but I don't >>> think that comes in at the normalization level. >>> >>> Sean >>> >>> On Oct 1, 2004, at 12:33 PM, Matthew Hannah wrote: >>> >>>> Hi, >>>> >>>> Quick question this time. Basically these dyes have higher >>> intensity, >>>> less quenching and correlate more between dye pairs. The ULYSIS kit >>>> uses chemical labelling which eliminates sequence effects. >>>> >>>> http://www.probes.com/servlets/directory?id1=6&id2=48&id3=319 >>>> http://www.probes.com/media/publications/394.pdf >>>> >>>> As far as I know Limma accounts for dye bias during >>> normalisation. If >>>> there is less/no bias will there be artifacts introduced? Or other >>>> points I need to be aware of? >>>> >>>> Thanks, >>>> Matt >>>> >>>> _______________________________________________ >>>> Bioconductor mailing list >>>> Bioconductor@stat.math.ethz.ch >>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>> >>> >>>