Calling copy number from SNP arrays

Question

[ethanb416]

I have log R ratios and beta allele frequencies from Affy6 data. I've used the LRR with the R package fastseg to segment the data. The problem is, I need copy number calls and it appears that fastseg just gave me segmentation means for the LRR and fancy plots. Can I use this to estimate copy number integers or do I need to start back at square one with my LRR & BAF using a different package? After segmenting with fastseg, I have ranges, num.mark (integer), seg.mean (numeric), startRow (integer), and endRow(integer). I've looked into PennCNV & cn.mops but I don't know much about them.

Any guidance would be greatly appreciated.

[karl.stamm]

Take a look at seg.mean, is it a lot of nearly two on autosomes, and one on sex-chr for males ?

[ethanb416]

So far I've only gotten to chromosome 1 but here's a summary of seg.mean

484 Celfiles

Min: -161.800

1st Qu.: -9.000

Median: -0.2105

Mean: -0.4894

3rd Qu.: 8.222

Max: 106.300

Answer 1

[Irsan]

It depends on your context. If you are working on tumor samples that most likely consist of a heterogeneous mix of cells, calling copy numbers in discrete integers doesn't reflect biology and you are better off with CBS implemented by e.g. DNAcopy/fastseg/CGHcall. When you work on germline CN[VA] calling then maybe it does make more sense and using hidden markov models that call copy numbers in discrete states might be more suitable.

[ethanb416]

Thanks, Irsan. I am working on germline data. Do you know if there is any preexisting R package that utilizes a HMM for that purpose?