need help with CNVs

[Robby]

Hello,

I have a lots of problems to call CNVs in genome samples (tumor/normal pairs) sequenced with Illumina.

First I tested CNVnator, but I had already problems with the second step, because I don't know the bin_size (./cnvnator -root file.root -chrom ch1 -his bin_size). I tested with 100, 1000, 1000000, but the program ran in all cases just for chromosome 22 over one week and was still not finished. I asked already in another thread for the bin_size, but unfortunally I didn't receive any answers (http://seqanswers.com/forums/showthr...t=11013&page=2).

After that I tested CNV-Seq, but I don't understand the output. Especially I don't know the exact position of the CNV and the difference between log2 and cnv.log2 and the difference between p.value and cnv.p.value

"chr1" 2250641 2250864 159 100 2250752 0.722575091988798 8.10063503680068e-07 77 448 0.964667045392917 3.93033619539229e-35
"chr1" 2250753 2250976 171 87 2250864 1.02845734551634 3.42115799964058e-11 77 448 0.964667045392917 3.93033619539229e-35
"chr1" 2250865 2251088 161 71 2250976 1.2347180850891 2.18142864565488e-14 77 448 0.964667045392917 3.93033619539229e-35
"chr1" 2250977 2251200 107 60 2251088 0.888124717271795 4.18272787914529e-09 77 448 0.964667045392917 3.93033619539229e-35

I tried to get some help in another thread (http://seqanswers.com/forums/showthread.php?t=16882) => unfortunally no answer.

Last I tried CNVer, but I can use that tool just for hg18, which doesn't really help me.

I there any other tool, which you can recommend? Or do you have any ideas/suggestions/help to solve the problems with one of the above mentioned tools?

I would be really happy about any comments, because I am completly frustrated.

Thanks a lot in advance.
Robby

[samesense]

Try out ExomeCNV. First, find LOH. Then use LOH regions to determine normal contamination in your tumor sample. Finally, estimate CNVs, using the normal contamination correction.

[Robby]

thanks for your answer.

But how do I search for LOH regions in tumor/normal samples? And does ExomeCNV also works for complete genome sequences or do I have to extract the exonic regions before?

Best regards
Robby

[samesense]

The program also does LOH. Follow these instructions. First, find normal tissue SNVs that are heterozygous (you have to determine these SNVs yourself), then make files with the coverage and 'mutant' read counts for these SNVs for tumor and normal tissue. These files are input for the ExomeCNV R LOH functions.

As for application to whole genomes, I'm not sure.

[clarissaboschi]

The bin size used depends on the coverage of your data, see supplemental material from CNVnator paper. But I have the same doubt. I tested bins of: 100, 120, 150 and 200 bp. My coverage is about 8-15X. With each bin I had a different number of CNVs. I am not sure which one is better...
Also I do not know if I should filter the results based on the p-value...