[BioC] Affymetrix preprocessing, MAS5 vs. RMA

[Kasper Daniel Hansen]

On Jan 22, 2009, at 13:00 , Sean Davis wrote:

> On Thu, Jan 22, 2009 at 3:47 PM, Dozmorov, Mikhail G. (HSC) <
> Mikhail-Dozmorov at ouhsc.edu> wrote:
>
>> Hello,
>>
>> I'm learning basics of importing Affy's .cel files to R, and do it  
>> with
>> affy library:
>>       R> Data <- ReadAffy()
>> When I do preprocessing
>>       R> eset <- rma(Data) OR eset <- mas5(Data)
>> I'm getting very different expression values.
>>       From MAS5:
>> ID      21.CEL  33.CEL  36.CEL
>> 1415670_at      2494.54 2576.29 2240.92
>> 1415671_at      6292.17 5013.13 3075.89
>> 1415672_at      6727.66 5858.31 5690.70
>> 1415673_at      567.38  718.02  875.25
>>       From RMA, Antilogged from Log2
>> ID      21.CEL  33.CEL  36.CEL
>> 1415670_at      74.16   69.30   59.08
>> 1415671_at      191.52  171.38  94.30
>> 1415672_at      194.18  199.57  175.13
>> 1415673_at      14.93   13.16   21.50
>>
>> Can't find any explanation why it could happen. And what to use?  
>> Please,
>> advice! Thank you. Mikhail.
>>
>
> Each normalization method is indeed different.  They should and will  
> produce
> different results.  There is an extensive literature on different  
> methods of
> normalization and preprocessing for Affy that you can read to get the
> details.
>
> I would say that most people would not use MAS5 anymore.  RMA is a  
> good
> alternative.

Also note that the actual numbers are meaningless. What you should  
compare is the (log) fold change between conditions. The actual scale  
has no meaning.

Kasper