[DOCKTESTERS] DKFZ BiasFilter 100% match on DO52140, DO35937, and DO218695

[Miguel Vazquez]

Dear all,

I'm very pleased to announce that with the help of Kortine we have managed
to reproduce the results. To get the proper results we had to filter out
the input file (consensus.vcf from GNOS for each donor) to remove the
 LOWSUPPORT and OXOG variants. I've also removed the bSeq and bPcr tags
from the file as I saw in the python code that otherwise these already
flagged variants would have been excluded.

Best regards

Miguel


Comparison for DO52140 tag bPcr
---
Common: 1
Extra: 0
Missing: 0


Comparison for DO52140 tag bSeq
---
Common: 1
Extra: 0
Missing: 0

Processing donor: DO35937

Comparison for DO35937 tag bPcr
---
Common: 1
Extra: 0
Missing: 0


Comparison for DO35937 tag bSeq
---
Common: 6
Extra: 0
Missing: 0

Processing donor: DO218695

Comparison for DO218695 tag bPcr
---
Common: 1
Extra: 0
Missing: 0


Comparison for DO218695 tag bSeq
---
Common: 10
Extra: 0
Missing: 0




On Wed, Mar 15, 2017 at 11:09 AM, Miguel Vazquez <miguel.vazquez at cnio.es>
wrote:

> Dear all,
>
> As you can read below I made a mistake on my previous validation for the
> DKFZ BiasFilter. Unfortunately large differences have turned up now that
> I've corrected the process. In brief on both donors I've found that
> re-runing the filter flags some additional variants for both flags bPcr and
> bSeq. Notably all the discrepancies are for the new method flagging more
> variants. For instance in many cases the original file contains just one
> variant with the flag where the new one ten or twenty. You can read the
> details at the end of this email where we are comparing the original VCF to
> the new one.
>
> Note that the orginal VCF is the consensus variants are the input I use
> for the BiasFilter along with the corresponding BAM files for that donor. I
> can only imagine that if this VCF was not the one originally used due to
> some filtering step then perhaps the bias calculations might have been
> affected. If that is so I would need instructions on where to get the
> precise input VCFs.
>
> Best regards
>
> Miguel
>
> ----RESULTS----
>
> Comparison for *DO52140* tag *bPcr*
> ---
> Common: *1*
> Extra: *12*
>     - Example: 11:81550771:C:A,12:19486241:G:T,2:12287406:G:T
> Missing: 0
>
>
> Comparison for *DO52140* tag *bSeq*
> ---
> Common: *1*
> Extra: *23*
>     - Example: 10:17681457:G:T,12:112049882:T:G,12:130990011:T:A
> Missing: 0
>
> Comparison for *DO35937* tag *bPcr*
> ---
> Common: *1*
> Extra: *10*
>     - Example: 1:114845662:G:T,14:33282600:C:A,16:78467879:G:T
> Missing: 0
>
>
> Comparison for *DO35937* tag *bSeq*
> ---
> Common: *6*
> Extra: *88*
>     - Example: 10:21703903:A:C,10:24183103:C:T,10:51468498:C:G
> Missing: 0
>
>
>
>
> On Mon, Mar 13, 2017 at 4:48 PM, Christina Yung <christina.yung at oicr.on.ca
> > wrote:
>
>> Hi Miguel,
>>
>> The bPCR and bSeq flags are indeed the ones flagged by the DKFZ bias
>> filter.  When you summarize the comparison, please cc Matthias of DKFZ as
>> his team developed this filter.  No issue at all, and thanks again for your
>> great work!
>>
>> Christina
>>
>>
>> On 3/13/2017 10:22 AM, Miguel Vazquez wrote:
>>
>> Dear all,
>>
>> I just learnt that the DKFZ BiasFilter is NOT the OXOG filter workflow,
>> which means* I checked for the wrong thing in this validation!* I'm
>> sorry for the confusion.
>>
>> Right now I pass the BAM files and the consensus.vcf (SNV_MNV) downloaded
>> from GNOS to the BiasFilter and compare the resulting VCF with the
>> consensus looking at the set of mutations containing the OXOGFAIL flag.
>> This apparently is not the comparison to make. *What is it that I need
>> to compare? is it the bPcr and bSeq flags?*
>>
>> One first look at those flags do show quite some discrepancies
>> unfortunately on both donors (DO52140 and DO35937) for both flags. For
>> instance for DO35937 we find 11 mutations flaged bPcr with in the new
>> result, while the consensus.vcf only finds one, of them. Something similar
>> happens with the bSeq.
>>
>> Can you please confirm this so I can come reply with a full report on
>> this.
>>
>> Kind regards, and sorry again for the confusion.
>>
>> Miguel
>>
>>
>>
>> On Mon, Feb 27, 2017 at 7:30 PM, Miguel Vazquez <miguel.vazquez at cnio.es>
>> wrote:
>>
>>> Dear friends,
>>>
>>> I've performed the first test with the DKFZ BiasFilter and got a perfect
>>> match. There are 55 variants annotated with OXOGFAIL and they are the same
>>> in the input VCF file (consensus SNV/MNV VCF for that donor) and the output
>>> of the BiasFilter. I'm running the test on a second donor.
>>>
>>> Best regards
>>>
>>> Miguel
>>>
>>
>>
>>
>> _______________________________________________
>> docktesters mailing listdocktesters at lists.icgc.orghttps://lists.icgc.org/mailman/listinfo/docktesters
>>
>>
>>
>> _______________________________________________
>> docktesters mailing list
>> docktesters at lists.icgc.org
>> https://lists.icgc.org/mailman/listinfo/docktesters
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.icgc.org/mailman/private/docktesters/attachments/20170327/5c062cbf/attachment.html>