CNVnator false positives -- Struggling

[dickie_ho]

Hey everyone

I am having a hard time dealing with CNVnator .. well the results it generates..
I have array data, exome-seq data which the CNV calls line up .... Now using WGS data and I am getting loads of calls (10,000s) from CNVnator but where I know there are CNVs (confirmed by taqman) I am not getting that called in CNVnator ...

My Pipe for WGS is bowtie2 -> Picard Mark dups -> GATK (base score and INDEL)

Appreciate any help as I have been stuck on this for weeks

Cheers

Josh

[Markiyan]

Did you use PCR-Free library preps?

Dear Dickie_ho,

So, which library prep technique did you (or your provider) use for wgs sample prep?
Nextera Shotgun, Truseq or PCR-Free?

For most reliable results we recommend using the PCR free one and aim for 350bp average fragment size (longer fragments are more prone to the CG bias in the clustering stage), and avoid the nextera with default cycling conditions on the fast ramping thermocycles.
The nextera's cycle denaturation times of 10-15sec are a bit to short, better use 30sec @96 or similar (PCR-free is still the best, esp @ high AT regions).

Otherwice I would not hope for much accuracy from the CNV detection based on the reads coverage.

Look on the reads coverge plots for the region where there is no CNV and the control ones, and get a feel for the SNR (signa to noise ratio).

PS: Lucigen recenly released a kit which is supposed to give you a PCR-free library with only 75ng of good quality input DNA (versus 1ug for illumina PCR-Free).

Markiyan.

[dickie_ho]

Hey Markiyan

Thank you for your reply

The TruSeq PCR-free preps were used

Will check out the SNR.. thank you for this suggestion.. I will have a look and see if I can some how determine the best calls.. if not back to the drawing board

####PS: Lucigen recenly released a kit which is supposed to give you a PCR-free library with only 75ng of good quality input DNA (versus 1ug for illumina PCR-Free).

Cheers for that info ..

Thanks

Josh

[Markiyan]

Also make sure the reads mapping to multiple regions are handled properly.

Dear Josh,

Last 5 cents:
Please also make sure the reads mapping to multiple locations are handled properly by the bowtie2. Maybe raise kmer length vallue to 25-31?
I would suggest mapping them randomly, do not just throw them out!
See CNVNator manual/guides for more info.

Optimal parameters would deppend on read length, read error rate and genome size/repetitiveness.

Markiyan.

[dickie_ho]

Found out from the sequencing centre that its was the nano not the PCR-free kit .....

Thank you for all your help Markiyan.. I would have spent a few more weeks trying to work this out if its wasn't for your help ....