[BioC] filtering

[J.delasHeras at ed.ac.uk]

I would only remove probes from the analysis if the signal is  
negligible in BOTH treatments, for ALL (or the majority) of the  
arrays. Otherwise you're likely to remove things that are expressed in  
one treatment but not the other and I can't imagine why you'd want to  
lose those, as it's an extreme case of up/downregulation.

With regards removing probes before of after normalisation... I  
personally prefer to remove them before. When I do filter (I don't do  
this always), I identify probes/spots with signal too close to  
background on each array, for each channel (mostly dealing with  
2-colour arrays myself). Then, as indicated earlier, I only filter out  
those that meet the criterion (for negligible  intensity) on BOTH  
channels (both comparisons, whatever), on every array (or 80% of teh  
arrays or whatever I deem appropriate in that particular case).

Then, those probes/spots are eliminated entirely (in limma, I just  
give them weight of zero), and work with the rest.

I don't think that removing them after normalisation makes much  
difference, in most cases, as they're going to be a bunch of spots on  
teh far left of the MA plot, converging towards zero if the background  
correction was good... and even if the background correction doesn't  
make teh low intensity spots converge towards M=0, in most cases they  
would be evenly distributed around M=0 (unless you have a very bad  
background problem) and the end result would be the same.

Jose



Quoting Lev Soinov <lev_embl1 at yahoo.co.uk>:

> Hi Jose,
>
>   I filtered those probes that had less than 50% of good signals in   
> at least one of the treatments. That was because I had four   
> treatments, making several contrasts in LIMMA. I did it in order to   
> make contrasts comparable. So, each contrast (pair of treatments)   
> contained the same set of probes. In your terminology I filtered on   
> one channel, keeping only signals detectable on ALL arrays.
>   The data shown in the plots are for just one of the arrays.
>
>   My question was about whether such filtering would affect   
> differential expression analysis if done before normalisation. Or it  
>  really does not matter when you filter before or after  
> normalization  step. Does this lead to anything undesirable or  
> wrong, apart from  loosing some genes that were "undetectable" in  
> one treatment but  "detectable" in some other treatments?
>
>   Thank you,
>   Lev.
>
>
>   J.delasHeras at ed.ac.uk wrote:
>   Quoting Lev Soinov :
>
>> Dear List,
>> I have posted a similar question before, but would like to ask you again
>> about filtering strategies. I have some AB1700 data and filter on signal to
>> noise ratios before normalization. The rationale is to get rid of badly
>> measured signals before actual processing of the data. Two jpg
>> histograms of
>> log2 signal distributions, before (raw.jpg) and after (filtered.jpg)
>> filtering, can be seen in this location:
>> http://tmgarden.cloud.prohosting.com/images/
>> Could you please have a look at the distributions and comment on whether
>> this is correct to filter before normalization as this changes the
>> distribution of
>> signals a lot?
>> Thank you very much for your help.
>> Lev.
>
> Hi Lev,
>
> Not sure what's the problem here. Your data has a large number of very
> low intensity points, and you removed them. The filtered histogram
> looks then exactly like the raw one, minus the low intensity stuff
> (the large peak on the left). That looks normal.
>
> I often do something similar, but I only remove the spots/probes that
> have low intensity in BOTH channels (if 2-colour hybs, if 1-colour
> then check on both samples that you are comparing), on ALL the
> relevant arrays. You seem to have filtered here only on one channel?
>
> Jose
>
>
> --
> Dr. Jose I. de las Heras Email: J.delasHeras at ed.ac.uk
> The Wellcome Trust Centre for Cell Biology Phone: +44 (0)131 6513374
> Institute for Cell & Molecular Biology Fax: +44 (0)131 6507360
> Swann Building, Mayfield Road
> University of Edinburgh
> Edinburgh EH9 3JR
> UK
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:   
> http://news.gmane.org/gmane.science.biology.informatics.conductor
>
>
>
> ---------------------------------
>  Yahoo! Answers - Get better answers from someone who knows. Tryit now.



-- 
Dr. Jose I. de las Heras                      Email: J.delasHeras at ed.ac.uk
The Wellcome Trust Centre for Cell Biology    Phone: +44 (0)131 6513374
Institute for Cell & Molecular Biology        Fax:   +44 (0)131 6507360
Swann Building, Mayfield Road
University of Edinburgh
Edinburgh EH9 3JR
UK