I am currently analyzing Copy Number Variant data from Affymetrix's Cytoscan HD platform on 45 cancer tumor samples. Normally, LOH and copy number variation would be looking at the tumor as compared with a control. Affy has told us that the control is unnecessary because their absolute values are accurate.
Given this information, how do I set thresholds for loss/gain based on probe intensity? I have looked in ChAS for this information, but am doing most of my analysis in Golden Helix SVS.
I can give more information if necessary, but wasn't sure what else someone might need to know to answer this.
Instead of a matched germ line control (which indeed is the best) you can use the median across arrays of your cohort as the baseline/control. The downside is that very frequent events could be misinterpreted. For example when 80% of your samples has 3 copies of chr1 and the rest has only 2 copies, 3 copies will be considered the baseline. Therefore, for 80% of your samples chr1 will look similar to the baseline (which is assumed to be 2 copies), while for the other samples it will look like they have a loss of 1 copy relative to the baseline. However, you will notice in your B-allele-frequencies. Although it is useful to know which SNPs are heterozygous in the matched germ line for LOH analysis, you don't need controls to calculate B-allele-frequencies.
Alternatively, you can use any other sample from public databases that was profiled using your array platform and use that as a control for all your samples in your cohort. Copy number polymorphisms (e.g. a gain of 50 kbps that is in the germ line and also in the tumor and therefore not acquired/somatic) will be identified as well.
You can do these analyses with for example Affymetrix power tools or aroma affymetrix.
Yes the more samples you have the better. But regardless of how big your cohort is, I think it is better to use an external cohort as baseline/controls. There are tons available.
Do you suggest using perhaps TCGA's controls for this? My concern is that I'm doing this in a non european-ancestry cohort. I worry that I may get some false results from using them. Any suggestions for where to find Cytoscan HD control data?
Just as an addition to this, a colleague mentioned that you need at least 25 samples to do this median value control.
Lesson: If Affy tells you that you don't need a germline control (and this will save you money), do it anyway.
Thank you.
Do you suggest using perhaps TCGA's controls for this? My concern is that I'm doing this in a non european-ancestry cohort. I worry that I may get some false results from using them. Any suggestions for where to find Cytoscan HD control data?