Question about control samples

[geekgeekhello]

Does ExomeDepth require multiple control samples for each run? Would having only 1 control sample work?

Thanks

[fbrundu]

I'm also interested. Following the vignette, using 1 tumour and 1 matched normal, I get no CNV calls. Correlation is above 0.97 and all the steps seem to work fine.

Edit: @geekgeekhello see section 11.1 of vignette

Edit 2: Also with somatic.CNV.call I don't get any call.

[vplagnol]

This is an old issue and probably not relevant anymore but I will attempt to reply nevertheless. In short, ExomeDepth is primarily designed for inherited changes. There was indeed a request to test matched pairs (tumour/control) and I implemented a small function to do this. However, this is rather ad-hoc and indeed one ends up quite limited by the size of the (typically) single control sample.

So yes, while large CNVs should be identified I would not be surprised if the power was quite low in that context, resulting in no calls quite often. I hope that answer makes sense?

I'll close the issue unless you guys have more comments. Apologies for the slow reply, my life hasn't had much time for ExomeDepth over the last year but I am looking into cleaning things up right now, and going over old issues.

[fbrundu]

@vplagnol Thanks for the reply. Yes it makes sense.

[laranonell]

Hi again,

this is also a question related to controls. I amb analyzing cancer panel data and I use panel design information to get the regions. I have just a few controls (16) and 110 samples. These samples do not have many alterations. I have created a reference with 15 of the 16 samples (following results from function select.reference.set) As I do not get expected results I think it might be a problem related to the controls. In your vignette/10. How to loop over the multiple samples, you use all files but the analyzed one to generate the reference set, do you think it is a better solution rather using the 15 control samples as a reference.set? Cheers,