search

bcbio / bcbio-nextgen

Validated, scalable, community developed variant calling, RNA-seq and small RNA analysis
https://bcbio-nextgen.readthedocs.io
MIT License
994 stars 354 forks source link

Tumour/Normal fields wrong way round for some calls in freebayes output #1848

Closed phu5ion closed 7 years ago

phu5ion commented 7 years ago

Hi,

While parsing a freebayes VCF file, I noticed that the last two fields on tumour/normal depth counts etc seemed to be the wrong way round, for some of the calls. I am sure that this is an error as I concurrently ran mutect and varscan and the calls were alright. For example, within the same vcf file: Chr Pos RefDepthNormal AltDepthNormal VAFNormal RefDepthTumour AltDepthTumour VAFTumour chr12 25245350 1410 412 0.226 1557 0 0 chr17 7674230 1286 1 0.001 830 1095 0.569

The chr12 variant is definitely the wrong way round and the chr17 variant is correct. Here's the vcf file, truncated to show only these two calls for brevity. I'm pretty sure it's not an issue with the Freebayes version as this issue didn't crop up when I last ran bcbio on the same Freebayes version. faulty_freebayes_output.zip

Thank you!

chapmanb commented 7 years ago

Thanks for the report. FreeBayes 1.1.0 did change the output order of multi-sample files. Previously they were in the order of the input on the commandline but that is no longer true with the latest versions. The latest release of bcbio (1.0.2) has fixes to handle this change.

However, I haven't seen inconsistent outputs between different called regions with FreeBayes. We do assume the sample order is consistent across all the regional VCF calls in a run. I'd definitely like to dig into this more and resolve the issue.

Could you provide the original regional VCFs called by FreeBayes for these two samples? These should be in freebayes/chr12/your_batch-chr12_start_end.vcf.gz. I'm trying to check if they have the same FreeBayes version and identify what would cause different sample orders. Thanks for the report and help debugging this.

phu5ion commented 7 years ago

Hi Brad,

I checked out the regional vcfs and indeed the sample order is inconsistent. freebayes.zip I think I might have restarted variant calling without removing intermediate files once, as I was running on a cluster and my allocated time ran out. Do you think this may have caused the issue?

chapmanb commented 7 years ago

Thanks much for the example and confirming the issue. The inconsistent outputs are with the same FreeBayes version and commandline calls. I pushed a fix to the latest development version which will force a consistent output order using bcftools view -s. After updating (bcbio_nextgen.py upgrade -u development). This will require re-running the freebayes part of your analysis to ensure they are all synchronized. I also filed an issue upstream to hopefully address the underlying issue in FreeBayes as well. Thank you for the help identifying the issue.