[BioC] RNA degradation problem

[James W. MacDonald]

fhong at salk.edu wrote:
> Hi Jim,
> Thank you for your useful input.
> 
> 
>>I find that the RNA degradation plots are less useful for indicating
>>possible problems than the density plots.
> 
> Do you mean that "RNA degradation plots are less useful for indicating
>  possible problems" or the problems indicated by RNA degradation plots are
> less profound than those from density plots?

Both. In my experience, if I see a sample that has high background 
(shifted to the right) on the density plot, it is almost always going to 
require a re-fragmentation and re-hyb. For the vast majority of samples, 
this is sufficient to get reasonable results. On the other hand, 
different slopes on the degradation plot often don't show up as being a 
problem (on residual plots, PCA plots, etc).

> 
> 
> 
>>If the density plots are all
>>reasonably similar, in my experience the normalization should be fine.
>>Another excellent plot for detecting problems is the residual plot in
>>the affyPLM package.
> 
> Two questions here, thanks.
> (1) Is it true that as long as the normalization is fine (for example, the
> boxplot and density plot after normalization are similar among arrays), we
> would proceed for further analysis.

Well, here is the problem. When you do a quantile normalization you will 
*always* end up with boxplots and density plots where the samples all 
look identical. That is pretty much what the normalization is supposed 
to do. The problem is whether or not it is a valid idea to use a 
quantile normalization for a particular data set (e.g., do the data meet 
the assumptions required to do a quantile normalization).


> (2) I did check the residual plot using affyPLM, P=please visit:
> http://cactus.salk.edu/temp/QC_t.doc
> What does it tell me? Some defects on the array?

There are some obvious defects on some of the arrays (my favorite being 
the 'window' on Stem_Base_1), but since the probes in each probeset are 
distributed all over the chip I don't think these will have a big 
effect. However, Stem_Top_2 has consistently large residuals, which 
indicates that the medianpolish fit does not explain these data very 
well. If we were running these chips in our core, this sample would 
almost certainly be redone.

HTH,

Jim


> 
> Thank you very much for your kind help.
> 
> Many thanks!
> Fangxin

-- 
James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623