[BioC] what to do with microarray outliers?

[Naomi Altman]

I have seen mislabelled samples.  Although either Wei or Jenny could 
be correct, tracking the samples all the way back to the person who 
took the tissue samples has paid off big for me.  (In one case, the 
person simply put cy3 and cy5 on the wrong columns of the lab book.)

--Naomi



At 12:36 AM 4/13/2011, Wei Shi wrote:
>Hi Theresa:
>
>         Although it is possible that samples were wrongly labelled, 
> I haven't encountered such a situation in all my microarray 
> analyses in the last several years. What more happened was that 
> arrays in the same experiment had variable qualities and in some 
> rare circumstances some arrays could have a really poor quality due 
> to for example sample preparation, hybridization problems and so 
> on. In these cases however it would be better to keep these arrays 
> in your analysis instead of throwing them away because these arrays 
> could still provide useful information. But these arrays should be 
> down weighted because their quality is poor. arrayWeights function 
> is designed to address this problem.
>
>Cheers,
>Wei
>
>
>On Apr 13, 2011, at 2:04 PM, Jenny Drnevich wrote:
>
> > Hi Theresa,
> >
> > If you have two groups that are clearly separated on a PCA plot, 
> except a couple samples are not in the "correct" group but instead 
> in the other group, then I'd wager my pension* that these are 
> mis-labeled samples, not outliers. I've seen this happen more than 
> once, and while disturbing, we usually can never figure out exactly 
> what happened. In general, outliers only refer to samples that are 
> different from ALL others, not just samples that cluster with 
> another group instead of their own.
> >
> > HTH,
> > Jenny
> >
> > * I work for the state of Illinois, so my pension may not be worth much :(
> >
> > At 11:06 PM 4/10/2011, Wei Shi wrote:
> >> Hi Theresa:
> >>
> >> Maybe you can try use array weights. See ?arrayWeights in limma 
> for more details.
> >>
> >> Cheers,
> >> Wei
> >>
> >> On Apr 10, 2011, at 9:56 PM, Theresa Brandt wrote:
> >>
> >> > Hello,
> >> >
> >> >   I would like to ask what to do if there are outlying 
> microarrays on PCA
> >> > plot. Should I remove them from the further analysis or not?
> >> >   I have two groups of samples and these groups are clearly separated on
> >> > the PCA plot (first and second PC). But a few microarrays are not in the
> >> > "correct" group.
> >> >   What to do in a situation that one of the samples is much 
> different from
> >> > all the others (but technically it as a good microarray)? What 
> would you do
> >> > in such a situation?
> >> >
> >> > Sincerely,
> >> > Theresa
> >> >
> >> >       [[alternative HTML version deleted]]
> >> >
> >> > _______________________________________________
> >> > Bioconductor mailing list
> >> > Bioconductor at r-project.org
> >> > https://stat.ethz.ch/mailman/listinfo/bioconductor
> >> > Search the archives: 
> http://news.gmane.org/gmane.science.biology.informatics.conductor
> >>
> >>
> >> ______________________________________________________________________
> >> The information in this email is confidential and intend...{{dropped:6}}
> >>
> >> _______________________________________________
> >> Bioconductor mailing list
> >> Bioconductor at r-project.org
> >> https://stat.ethz.ch/mailman/listinfo/bioconductor
> >> Search the archives: 
> http://news.gmane.org/gmane.science.biology.informatics.conductor
> >
>
>
>______________________________________________________________________
>The information in this email is confidential and inten...{{dropped:11}}