[BioC] snapCGH and CGH method evaluation

[Sean Davis]

João Fadista wrote:
> Dear all,
>
> snapCGH -> Does anyone know how to display regions of gain or loss across multiple samples in the same plot? I am asking this with the purpose of finding common breakpoints and regions of common gain or loss between samples.
>
>   
The "standard" way of doing this is a "frequency plot".  Simply sum 
across samples all the probes that are greater (less) than a certain 
threshold and plot them along the chromosome.  I don't personally like 
them very much, but it is easy to produce.  Also, you might look at the 
package cghMCR and also at the java executable called STAC.

> CGH method evaluation -> Does anyone knows if there is any CGH package that estimate the statistical significance of the detected copy number changes and then rank them accordingly?
>   
This is a very challenging problem.  You, as the researcher, will 
probably need to decide what you think is important (short segments, 
long segments, high-copy number gains, homozygous deletions, copy number 
polymorphisms, etc.) and design your own methods for pulling these out.  
A simple statistical test does not do the trick, in my experience, for 
pulling out all the relevant information.  If you really want something 
that gives statistical significance, you might want to look at an 
article by Lipson et al., in RECOMB 2005 (don't have the exact reference 
at hand). 

Sean