[BioC] arrayQualityMetrics

[audrey at ebi.ac.uk]

Hi Steve,

If we consider an NChannelSet named obj and the pData of obj look like this:

      Treatment  Cy3   Cy5   Replicate
Array1         U   WT   Ref           1
Array2         S   WT   Ref           2
Array3         U   MU   Ref           1
Array4         S   MU   Ref           2
Array5         S   MU   Ref           1

If in your quality report you are interested in seeing the heatmap with a
side colour bar representing the treatment, the right way to call the
function is:
arrayQualityMetrics(obj, intgroup = "Treatment")

If you want to see the heatmap with the colour bar representing the
replicates, you need to use:
arrayQualityMetrics(obj, intgroup = "Replicate")

The default column name of the phenoData that would be used for the colour
side bar is "Covariate", but you can use any column name of your
phenoData.

I do not know what is in your p object but you can use any of the column
name of p that makes sense for you to be represented next to the heatmap.
This is completely optionnal and you can perfectly run the
arrayQualityMetrics function without setting an group of interest.

I hope that helps,
Audrey


> Hi Audrey,
>
> Thanks for your reply.
>
>>
>> 1) The absence of background intensity will not have any hidden effect
>> on
>> the report. The only plot of the report that would make use of the
>> background intensity is the spatial distribution, to see spatial
>> effects.
>> If you have R, G, Rb and Gb, each of them will be represented (4 spatial
>> plots per array) and if only R and G are available, only 2 spatial plots
>> will be represented per array. There is no processing within the
>> arrayQualityMetrics function. If you want to subtract background or
>> normalise your data, you need to do it before (using limma for
>> instance).
>>
>
> ok.
>
>> 2) The RIN column is not needed in the phenoData, it was just an
>> example.
>> Another example would be, if you have a factor of interest, like
>> treatment/control you can define a column in your phenoData with this
>> information and then, when you call arrayQualityMetrics you can set the
>> argument intgroup (which stands for interesting group) equal to the name
>> of the column containing the treatment/control information. This will
>> draw
>> a colour side bar to your heatmap.
>
>
> I am currently doing
>
>  > p=read.AnnotatedDataFrame('sinfo.txt')
>  > varMetadata(p)$channel=factor(c("G", "R", "G",
> "R"),levels=c(ls(assayData), "_ALL_"))
>
> Can you send me an example of what you mean?
>
> Many thanks,
>
> Steve
>