I am doing some CNV calling analysis on whole Exome sequencing data. Can anyone recommend some tools for CNV on WES data?
I have done some literature search, and VarScan2 is one of tools suitable for this type of data, but does VarScan2 have to take disease-normal paired data? Can I run VarScan2 on normal and disease, respectively, and merge the results at a later step? I know this would not make much difference in terms of the disease-specific CNVs identified, but I am trying to fit the CNV calling to a pipeline, and I need to stick on the ways the existing pipeline works.
Recommendation on other tools are welcome.
Thanks!
Hi Jia,
I've tried most of the popular tools including Varscan2, ExomeCopy, FREEC and a few others.
http://www.bioconductor.org/packages/release/bioc/html/exomeCopy.html
http://seqanswers.com/forums/showthread.php?t=16803
http://varscan.sourceforge.net/
They all claim to work well in their respective papers - often using simulated data - but I've found the results in real life to be highly variable and sensitive to sample quality and the amount of sequence in each sample. I've ended up with the following pipeline that I'm pretty happy with. I setted on the recent version of FRECC within the following steps:
1) Align with BWA as standard
2) If you don't have matched normal sequence, then make a pooled normal sample by combining all of your samples then downsampling (Not pretty but it works).
3) Balance your normal and control samples by down sampling whichever is biggest. I've found a big size differential causes problems with most of these tools.
4) Run FREEC in exome mode as detailed on their website.
5) Ignore the overly sensitive HMM that they use to segment.
6) Re-segment with DNAcopy
Happy to share the specifics if you're interested.
Hi Jia,
As is clear from the earlier answers there is no one tool that will do the job. Some tools are better at detecting small CNVs while others excel only in detecting rare events, also see this thread on biostar: https://www.biostars.org/p/13807/
About two years back we reviewed several Exome specific tools for their poer to detect known clinically relevant CNVs in comparison to a high resolution aCGH platform (see link 1). Since then many new tools have been developed and tested, each performing better in a particular area. Look up a recent review (such as link 2 and 3) to get the ins and outs.
If you're looking for tools that are modified for pipeline usage take a look at the tools/pipeline developed by the Sanger https://github.com/cancerit/. cgpBattenberg is propably the one you would be interested in for this question
The key to almost all tools is a high quality control sample or pool of samples that capture the enrichment/sequencing biases of your experimental setup. The question which tool will then best suit your needs depends on several factors like if you want to detect rare events (likely a Z-Score based approach) or shared events (more likely a HMM). And also on if you have a specific application in mind, e.g. tools optimised for cancer samples usually assume a single reference sample while more general tools usually require a pool.
Hope this helps in your decision making.
Best,
Joep
http://www.ncbi.nlm.nih.gov/pubmed/23893877
http://www.ncbi.nlm.nih.gov/pubmed/24599517
http://www.ncbi.nlm.nih.gov/pubmed/25102989
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools