CNV analysis tool Question With two samples tumor and normal,is there any software for CNV in exome or whole genome sequencing

Question

[zhaolianhe1992]

Hi all，

My data is two samples: tumor and normal, exome and whole genome sequencing data.

I have tried XHMM for CNV analysis, but it needs more than 2 samples, so I changed to CNVnator and CNVFinder, but they are very hard to install in my system since I do not have root access and centos needs many other libraies which are not installed in my system.

I used CNVseq too, but the outcome is not very good.

I used varscan2 to analyze CNV, but the outcome is not very good either.

So is there any other CNVtools for tumor-normal sample?

Many thanks!

lianhe

[Chirag Nepal]

I am using varscan2 for exome CNV, and the result looks fine.

Can you tell/show why your result is not fine ?

[zhaolianhe1992]

Hi Chirag:

Thank you for reply!

Can I ask you how you analyze CNV for exome by varscan2 in detail?

When I use varscan2,I followed the recommonded pipeline in the varscan2 official website,which is:

1. Run VarScan copynumber on normal and tumor mpileup output

samtools mpileup -q 1 -f ref.fa normal.bam tumor.bam | 
  java -jar VarScan.jar copynumber varScan --mpileup 1

This will create a single output file, varScan.copynumber, containing the raw copynumber calls.

2. Run VarScan copyCaller to adjust for GC content and make preliminary calls.

java -jar VarScan.jar copyCaller varScan.copynumber --output-file varScan.copynumber.called
 [--output-homdel-file varScan.copynumber.called.homdel]

This will create two output files: varScan.copynumber.called (adjusted calls) and varScan.copynumber.called.gc (GC adjustment information). As of version 2.2.12 (August 2012), you can also specify an output file for candidate homozygous deletions.

After the first step, I get output.copynumber file like:

chrom    chr_start    chr_stop    num_positions    normal_depth    tumor_depth    log2_ratio    gc_content
chrM    1    100    100    14.7    5.1    -1.537    0.0
chrM    101    200    100    16.7    7.9    -1.075    0.0

I do not know why all the gc_content is 0.

After step 2, I got:

chrom    chr_start    chr_stop    num_positions    normal_depth    tumor_depth    adjusted_log_ratio    gc_content    region_call    raw_ratio
chrM    342    441    100    37.5    45.8    0.289    0.0    amp    0.289

well, in my opinion,I think varscan2 just used the ratio between tumor and normal, and cut the threshold to identity them as amp or deletion or neutral. I do not think this is a good way to analyze CNV.

Maybe I do not know varscan2 very well, so please give me some advice.

Many thanks!

lianhe

Answer 1

[Eric T.]

Yes, many. CNVkit is a tool that you haven't mentioned but should work for your project. (Disclosure: I'm the author.)

[Chirag Nepal]

Hi Etal,

Can you use CNVkit if you want to compare the copy number analysis of N number of T/N exome pair. Something like what J/GISTIC does ? If CNVkit can be used to get the combined CNV analysis, could you please share how it can be done.

Thank you !! Looks like a nice tool. Will definitely try.

[Eric T.]

Sorry I missed this earlier. CNVkit doesn't do a cohort statistical analysis itself; it mainly delivers the segmented log2 ratios for each sample, optionally rounding to integer copy numbers, and provides a number of ways to feed that data to other programs for other analyses. I think GISTIC takes the SEG format as input, so you can use CNVkit to export the segments from all your samples to SEG and feed that to GISTIC for the cohort analysis (for example).

[Chirag Nepal]

Thanks Etal !!

I ended up using varscan2/GISTIC2. I have posted a tutorial here, hopefully will be useful to some.

Tutorial: Analyze exome Copy number variation (CNV) in single patient or in population.