[BioC] Differential Expression Methods

[Darlene Goldstein]

hi.....

martin doherty <martin.doherty at ...> writes:

> 
> Hi,
> 
> I have Tumour and Normal technical replicates run on both FFPE and
> RNA-Later.
> These are run on X3P arrays. I am interested in comparing the outcome from
> both
> fixation methods. When I compare the transcripts classed
> as expressed in both RNA-Later and FFPE the ones found
> unique to FFPE is very small (<5%) - if I assume the RNA-Later
> samples exhibit true expression levels this is a good result.
> The intensities/background from both fixation methods is almost
> equal as well.
> 
> When I look for differential expression between tumour/normal using
> both fixation methods. I find that almost half of what is differentially
> expressed in FFPE is not differentially expressed in RNA-Later. This
> disagreement would indicate that the differential expression in FFPE
> is peppered with false-positves or that RNA-Later is not picking up
> differential expression - I'd guess it is the former.

I don't have specific experience with microarrays on this, but have looked at
differences between these using PCR - see J. Antonov et al, 2005; 
Gene Expression Measurements by Quantitative Real-time PCR Depends on Short
Amplicons and a Proper Normalization; Laboratory Investigation 85:1040-1050.

It would also be my guess that you are getting false positives in the ffpe
samples, probably from differing efficiency of hybridization in the different
ffpe samples (again, just a guess).  You could try to assess this by looking at
how variable the different ffpe hybs are (within condition) and compare that to
the same thing in the rnalater samples.

Best regards, Darlene

> 
> The tests for differential expression was done using CyberT, SAMR and
> Muilttest,
> and a modified t-test. Does any one have any ideas what other testing
> methods would
> be worth investigating, or have any idea why there is so many false
> positives
> in FFPE?
> 
> Thanks,
> 
> Martin
> 

--
Darlene Goldstein
École Polytechnique Fédérale de Lausanne (EPFL)
Institut de mathématiques
Bâtiment MA, Station 8        Tel: +41 21 693 5565 
CH-1015 Lausanne              Fax: +41 21 693 4303
SWITZERLAND