[BioC] VSN with spike-in probes question

[Wolfgang Huber]

Dear Matthew

it makes sense to subset the spike-in controls used for normalisation with vsn2 to those that respond about linearly to changes in the target mRNA abundance.

lts.quantile=1 is reasonable choice for such a situation. The ordering of the features makes no difference to vsn2.

Kind regards
Wolfgang





On 24 Jun 2014, at 21:36, Thornton, Matthew <Matthew.Thornton at med.usc.edu> wrote:

> Hello!
> 
> I am trying to optimize my data processing based on the addition of ExFold ERCC controls. Ideally I would like to normalize with VSN using the procedure in chapter 7 of the vsn vignette. If I pull out the unprocessed intensities for the ERCC controls, order them by increasing concentration, and  transform the concentrations and intensities by log base 2. I get a sigmoid curve which I can fit.  Should I only use the probes which show a linear relationship as the spike-ins with lts.quantile=1? Should I order the "spikeins" by increasing concentration before passing to vsn2? I think that a significant source of error are intensities set as constant outside the dynamic range of the array. For example the intensity of the lowest 5 concentrations oscillate around log2(intensity)=5. We see a linear range in log base 2 intensity from 6.0-6.5 to 12. above 12 I see saturation. Also if it is only important that the "spike-in" probes are of similar intensities ("not expected to be diffe!
> rent") should I use the hybridization controls or poly-A controls that are above log2(intensity) of six.  Any advice or assistance is greatly appreciated.
> 
> Thanks,
> 
> Matt
> 
> 
> 
> matthew.thornton at med.usc.edu
> 
> _______________________________________________
> 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