[BioC] array CGH - raw data analysis

[Sean Davis]

On Tue, Mar 4, 2008 at 1:36 PM, João Fadista <Joao.Fadista at agrsci.dk> wrote:
> Dear all,
>
>  I have a question related to the possibility of analysis of raw intensity data of array CGH experiments.
>
>  After I finished my analysis of 2-color array CGH data with the usual "standard" protocol (transform raw intensities in log2ratios, normalize, segment the data and call CNVs with some chosen criteria) I wanted to know the true copy number status of my CNVs, without further experiments. Every array CGH study reports the status of copy number variations in relation to the reference sample but is it possible to know the true copy number status?
>
>   I tried three inconclusive approaches:
>
>  1º - Since my experiment has 24 arrays with the same common reference (in Cy5), I averaged the Cy5 raw intensities of the probes for these 24 arrays and then plotted them across the genomic regions queried. When looking at the plots I tried to see if, for the CNV regions found, there is a significative increase or decrease in the value of the intensities of the probes compared to the neighbouring normal regions.
>
>  2º - Segmented the raw intensities (both test Cy3 and Cy5 reference samples) instead of segmenting the normalized log2ratios. The segments found in both kind of samples were not the same as in the "standard" protocol.
>
>  3º - Made t-tests and Mann-Whitney tests comparing the distribution of the raw intensities (both Cy5 and Cy3) of the CNV regions with the normal copy number regions immediately upstream and downstream of them.
>
>  Is this quest for assessing the true copy number status a lost case?

Hi, Joao.  Comparing intensities from one probe to the next is not
likely to be terribly fruitful (at least not without some more
processing), I do not think.  Also, keep in mind that CNVs in ratio
space are due to variation in both the reference and the sample.  As
for the absolute copy number, you didn't mention what type of samples
you are dealing with, so there may not even be an integer absolute
copy number associated with every point in the genome due to
heterogeneity.

Sean