Tools For Analyzing Copy Number Variation

Question

[1036268670]

Dear all:

I am doing a project about detecting copy number alterations. I have the exome sequencing data of 9 cell lines after anti-tumor drug treating:8 bladder cells (5637 bladder cancer cells, BIU87 bladder cancer cells, EJ bladder cancer cells, Hbc bladder cancer cells, J82 bladder cancer cells,T24 bladder cancer cells, UM_UC_3 bladder cancer cells) ,and 1 normal cell line (human uroepithelial SV-HUC-1 cells). Among them, 3 cells are drug resistant (SV-HUC-1,5637,and J82), other 6 cells are drug sensitive. I want to find some drug-resistance associated genes with copy number alternation. I learn about some CNV tools need paired samples. I don't know whether the 6 drug sensitive cells can be used as control .Which CNV tool are suitable for my problem? Thank you very much.

Answer 1

[skbrimer]

I am not 100% sure since I do not do the type of analysis that you are asking about, however I would imagine that there are many. Here is a link from this forum on CNV tools. If you are familiar with the R language I know that Bioconductor does this well.

For your question on which should be the control or not, I'm not sure on that one. It sounds like you are just looking at everything so I would imagine that there is not a real control in this. It is a comparison between normal, drug sensitive, and drug resistant. So assuming that your experiment is "let us sequences these known phenotypes and see where they differ to discovery areas of interest", then you are looking for areas that differ from the norm and from each phenotype. i.e normal cells have 2 copies of gene x, DS have 10 copies of gene x, and DR have 0 copies gene x.

I'm sorry I'm not more helpful, but maybe this will get you on the right track.

[1036268670]

Thanks for your reply. The CNV tools your provided me is very helpful. I am familiar with the R language, so I can try it. Thanks again.

Answer 2

[Eric T.]

Several callers including cn.MOPS and CNVnator (both good and relevant for your project) will call CNVs in samples without a matched control. If you're looking for high-level amplifications of full genes then reference-free calling from on-target read depth might be fine.

If you want more control over how you build your reference, CNVkit (mine, sorry) will also let you do control-free calling on each of your samples. Then, if you like, you can manually examine the results build a reference profile from either a subset of the original samples that have fewer recurrent CNVs, and/or manually zero out the detected CNVs in the pooled reference you've built from those samples so that calling in just those regions will avoid a false negative for recurrent CNVs.