Any better ways to filter GATK Mutect2 output?

Question

[ShirleyDai]

Hi, I tried GATK Mutect2 with default parameter, as follows:

java -jar GenomeAnalysisTK.jar \
     -T MuTect2 \
     -R reference.fasta \
     -I:tumor tumor.bam \
     -I:normal normal.bam \
     [--dbsnp dbSNP.vcf] \
     [--cosmic COSMIC.vcf] \
     [-L targets.interval_list] \
     -o output.vcf

But I got hundreds of mutations for one sample. I wonder if is there a referenced filtration step that can be used. Thanks

Answer 1

[Zhenyu Zhang]

hundreds of mutations may not be an indicator of bad data. Have you tried other callers and compare results together?

[mforde84]

agreed. when people say stuff like the exonic mutation rate is on average 1 per kb, they are talking about germline calls. so if your looking at wgs or exome, then yea you might see 10s of thousands of events (the majority of which are probably non-synonymous and or high MAF). frequencies of somatic mutations are an entirely different story.

[ShirleyDai]

Mutect2 can be only used for somatic mutation calling, not germline. I have not try other tools yet.

Answer 2

[igor]

Can you provide more context? Hundreds of mutations could be too few or too many. What is the target region size and average coverage?

According to GATK:

Early validation results suggest that MuTect2 has a tendency to generate more false positives as compared to the original MuTect; for example, it seems to overcall somatic mutations at low allele frequencies, so for now we recommend applying post-processing filters, e.g. by hard-filtering calls with low minor allele frequencies.

Source: http://gatkforums.broadinstitute.org/gatk/discussion/6495/version-highlights-for-gatk-version-3-5

[ShirleyDai]

Thanks. MAF filtering might be helpful. I'm calling somatic mutation from whole exome sequencing data of skin cancer.