[BioC] kooperberg correction

[Matthew Ritchie]

Hi Jason,

> Hi Matt
>
> One of my gpr files has data from a
> top row of a sub grid with values of 0
>
> looking at the original image these spots have been printed right on the
>  edge of the possible scan area
> where I'm guessing information is being lost when genepix is extracting
> data.
>
> Is there an arbitrary number I could give this small set of spots to
> replace the 0's

Yes, the easiest thing to do is to make up some values for those spots to
keep the kooperberg() function happy.  I'd make a temporary .gpr file and
replace the relevant rows with complete rows from nearby spots.

> or would that defeat the object

It won't make a lot of difference to the model, as the adaptive background
correction is done on each channel separately for each spot (there are a
few shared parameters estimated for each channel across the whole array,
however these are unlikely to change).  You can reset the R, G values for
the lost spots to missing once kooperberg() has finished.

Sorry about the messy solution.  Best wishes,

Matt Ritchie

> I've tried the kooperberg approach on other gpr files and they all work
> fine
>
> many thanks
>
> Jason
>
>>This means that one of the limits (lower/upper) passed to the function
>> integrate() is a missing value (NA) for one of the genes.  I wouldn't
>> have expected this error because the limits are calculated using data
>> from the GenePix results files without logging.
>>
>>Can you check that the columns "F532 Mean", "F532 SD", "B532 Mean",
>> "B532 SD", "F635 Mean", "F635 SD", "B635 Mean", "B635 SD", "F Pixels"
>> and "B Pixels" in your files don't have any missing values?
>>
>>If there are no NA's, perhaps you can send me an example of one of your
>> .gpr files so I can take a closer look at the problem.  Best wishes,