[BioC] A CGH platform question

[Henrik Bengtsson]

Just a few general comments without giving you a direct answer:

The Illumina and Affymetrix platforms also provides genotypes for the
SNP loci, the Agilent platform doesn't.  If you at some stage want to
look at genotypes or LOH studies as well, you cannot work with a pure
CGH platform such as Agilent.

Also, there is information in the genotypes on the overall CN (e.g.
allele frequencies).  In the best of worlds (when
calibration/normalization is perfect) this information is orthogonal
(supplementary) to the CN signals, but this is often not the case.
Then the distribution of allele ratios in a CN region will help you
conclude what the total CN is in that region, e.g. for a CN=2 region
you would expect to see allele-specific CNs on bands of 0, 1, 2 copies
of allele B.  For a deletion CN=1, you will only see bands of 0 and 1
copy of allele B.  For a gain, you'll see bands of 0, 1, 2, 3 copies
and so.  This is well known and there are a few publications on this.
I think there is also work on how to improve CN calls given this
information, especially if you have homogeneous samples.  So,
potentially, I believe you have extra power calling CN states with SNP
arrays.

The different platforms have different coverage in different regions,
so you want to check how many loci each platform has in the regions
that are of special interest to you.  When you say "CNVs", do you mean
(i) CN aberrations (in single samples), or (ii) CN polymorphism (CNP;
across multiple samples)?  That's two different objectives.  I know
that at least Affymetrix designed its GenomeWideSNP_6 array to target
known CN polymorphic regions.  Don't know the details on Illumina, and
I think the Agilent 244K design predates the published CNPs results
(correct me if I'm wrong).

Finally, don't get fooled by the apparently different noise levels of
the different platforms when you look at the full resolution.  Things
that matters are their coverage in regions that you are interested in
and how great there *effective* resolution is in those regions (i.e.
how well can they detect a CN changepoint and at what precision - if
that is of interest to you).

Cheers

Henrik

On Wed, Dec 17, 2008 at 3:54 PM, Johan Lindberg <johlin at kth.se> wrote:
> Dear All.
>
> I am starting up a new project in which we are going to do CGH on a
> large set of paired samples. I have not decided on which platform I
> should use to do this though. The competitors are Agilent 244k, Affy
> SNP 6.0, Illumina 1M.
>
> I have looked in the recent nature and nature genetics papers where
> 76 - 27 % of the CNVs detected by a fosmid approach (Mapping and
> sequencing of structural variation from eight human genomes) were
> verified depending on CNV size using affy SNP 6.0 (Integrated
> detection and population-genetic analysis of SNPs and copy number
> variation).
>
> The sensitivity of Illumina 1M was 47% in another paper (Systematic
> assessment of copy number variant detection via genome-wide SNP
> genotyping) compared to again the fosmid approach.
>
> I havent seen any data on the Agilent array for these patients so its
> hard to say anything on this one.
>
> I would greatly appreciate if someone with experience of CGH from
> could comment on which platform that would give the best coverage of
> CNVs and if there are any other reasons to why one should prefer one
> platform over the other.
>
> best regards,
>
> // Johan
>
>
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> 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
>