[BioC] aCGH

[Sean Davis]

On Wednesday 29 November 2006 05:26, João Fadista wrote:
> Hi Sean,
>
> It seems that I have successfully made the log2.ratios and the clones.info
> files ready to start analysis with the aCGH package. But I have 2
> questions:
>
>
> 1 - When I use the impute.lowess function it give me some warnings. Do you
> know if there is any problem? I put the clones.info and log2.ratios files
> in attachment. Changing the argument maxChrom = 23 it gives the same
> warnings.
>
> > log2.ratios.imputed(ex.acgh) <- impute.lowess(ex.acgh, maxChrom = 24)
>
> Processing chromosome  1
> Processing chromosome  2
> Processing chromosome  3
> Processing chromosome  4
> Processing chromosome  5
> Processing chromosome  6
> Processing chromosome  7
> Processing chromosome  8
> Processing chromosome  9
> Processing chromosome  10
> Processing chromosome  11
> Processing chromosome  12
> Processing chromosome  13
> Processing chromosome  14
> Processing chromosome  15
> Processing chromosome  16
> Processing chromosome  17
> Processing chromosome  18
> Processing chromosome  19
> Processing chromosome  20
> Processing chromosome  21
> Processing chromosome  22
> Processing chromosome  23
> There were 50 or more warnings (use warnings() to see the first 50)
>
> > warnings()
>
> Warning messages:
> 1: collapsing to unique 'x' values in: approx(lowess(kbl[ind], vecl[ind], f
> = smooth), xout = kbl[-ind]) 2: collapsing to unique 'x' values in:
> approx(lowess(kbr[ind], vecr[ind], f = smooth), xout = kbr[-ind]) .

Do you have multiple probes with the same chromosome locations?  I suspect 
yes.  I think these might be responsible for the warnings, but I'm not sure.  
I would simply plot the non-imputed data and the imputed data to make sure 
that the imputation is working as you think it should.

> 2 - When I plot my acgh object it give me some strange plot comparing to
> the examples in the aCGH library. I put also this file in attachment.
> Thanks once again.

What is "strange" about it?  Keep in mind that often CGH segmentation methods 
need some tuning to get the performance correct for a given dataset.  It may 
be that you need to do that for your data.  

Sean