[BioC] cellHTS: help for NPI normalization

[Florian Hahne]

Hi Klaus-Peter,
taking a look at the source code for function NPI it seems that it 
doesn't support activator and inhibitor controls, which I guess makes 
sense. NPI is defined as

(my_pos - x_i) / (my_pos - my_neg)

and assuming that the set of positive activator controls has positive 
values and the inhibitor controls have negative values, trying to 
combine them doesn't give you anything useful. You could use either the 
inhibitor or the activator controls like this:

x <- normalizePlates(x, normalizationMethod="NPI",negControls=negCtr, 
posControls="(?i)^Mock$",transform=log)

Alternatively, you could treat your inhibitor controls as "negative" 
controls, since all you want to archive with the normalization it to map 
your values into a range that is specified by any of the controls you 
are using. Technically, this wouldn't be a "normalized percent 
inhibition", but then again inhibition is not what you are looking for 
exclusively in a two way assay...

Hope that helps, maybe Wolfgang or Greg have something to add to this.
Florian


Klaus-Peter Pleissner wrote:
> Dear all,
>
> using cellHTS (not cellHTS2) we want to perform a "NPI-normalization" 
> with positive and negative control wells.
>
> According twoway. pdf vignette we define  control wells  as :
>
> posCtr <- list(act= "(?i)^Mock$", inh = "(?i)^NP$")
> negCtr <- "(?i)^Allstars$"
>
> and normalize by
>
> x <- normalizePlates(x, normalizationMethod="NPI",negControls=negCtr, 
> posControls=posCtr,transform=log)
>
>
>
> We get an error:
>
> Error in NPI(x, posControls, negControls) :
>  'posControls' should be a vector of regular expressions with length 1
>
> Enter a frame number, or 0 to exit 
> 1: normalizePlates(x, normalizationMethod = "NPI", negControls = 
> negCtr, posCo
> 2: switch(normalizationMethod, mean = scaleByPlateMean(x), median = 
> scaleByPla
> 3: NPI(x, posControls, negControls)
>
>
>
> If we are using  a negative normalization  method ( as descibed in the 
> vignette)
> x <- 
> normalizePlates(x,normalizationMethod="negatives",negControls=negCtr, 
> posControls=posCtr,transform=log)
>
> this error doesn't occur.
>
>
> What is wrong in the definition of positive control wells ?
> Thanks is advance for help.
>
> Klaus-Peter


-- 
Florian Hahne, PhD
Computational Biology Program
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M2-B876
PO Box 19024
Seattle, Washington 98109-1024
206-667-3148
fhahne at fhcrc.org