How do I set less stringent germline filtering in Mutect2 / filtermutectcalls, many false negatives
Answered
Multiple mutations are filtered out as germline by filtermutectcalls, including very specific tumor hotspot mutations, often those mutations are located at a gain and have therefore a higher AF than the other mutations, is there a way to make the germline filter more stringent? Or another way to prevent that Mutect2 filter those out?
I did read the Mutect2 documentation, and saw the complex statistical approach of the germline filter, but could not find any settings that would make the filtering less stringent.
Because it is very important for the analysis to not miss those events I can think of 2 sollutions: 1- keep all mutations marked as germline in further downstream analysis – what will result in false positives
Or
2 – use a database like clinvar, cosmic, cancer gene census to at least keep all cancer mutations which were marked as germline.
But both solutions are workarounds and therefore not preferable, but is there another solution?
I’m running Mutect2 (4.1.7.0) as follows
gatk Mutect2 -R hg19.ref -I {exonbam -O output.raw -germline-resource gnomad} -pon {params.pon} -L intervalfile --f1r2-tar-gz output.f1r2 --max-mnp-distance 0 --native-pair-hmm-threads {threads}
gatk LearnReadOrientationModel -I input.f1r2 -O obmodel
gatk GetPileupSummaries -I exonbam -O pileup -V variants -L interval
gatk CalculateContamination -I pileup -tumor-segmentation output.segments -O output.contamination
gatk FilterMutectCalls -V input.raw -R hg19.ref --tumor-segmentation input.segments \
--max-events-in-region 50 --min-allele-fraction 0.05 \
--min-reads-per-strand 3 --unique-alt-read-count 4 \
--contamination-table input.contamination --ob-priors obmodel -O output.filtered
A line that was filtered for example is a MYD88 hotspot mutation:
Mutect2_filt.vcf: chr3 38182641 . T C . germline AS_FilterStatus=SITE;AS_SB_TABLE=24,17|20,17;DP=81;ECNT=1;GERMQ=1;MBQ=20,20;MFRL=167,158;MMQ=60,60;MPOS=23;POPAF=4.44;ROQ=82;TLOD=81.86 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:41,37:0.468:78:29,13:12,23:24,17,20,17
I hope somebody can help me.
Kind regards
Tjitske
If not an error, choose a category for your question(REQUIRED):
a)How do I (......)?
-
Hi Tjitske de Vries,
Tumor-only calling is very challenging when it comes to variant filtering. Having lenient germline filters will end up with a high percentage of sites in the output being germline sites. However, like you said, having more stringent germline filtering can throw out real sites.
You can use the option --f-score-beta in FilterMutectCalls to adjust the balance of the germline filters. We don't have a lot of documentation yet on how to adjust this parameter so if you find any improvements, please post here since it might be helpful for other users running tumor-only calling. The only note is that it cannot be negative.
Hope this helps!
Genevieve
Please sign in to leave a comment.
1 comment