SOS -- MuTect is picking up obvious germline mutations

Question

[haiying.kong]

This is the command I used to run MuTect:

/usr/bin/java -Xmx4g -Djava.io.tmpdir=tmp -jar ${MuTect}mutect-1.1.7.jar \
--input_file:normal ${BQSR_dir}${Blood}.recal.bam --input_file:tumor ${BQSR_dir}${Nevus}.recal.bam \
--out ${MuTect_dir}${Blood}${Nevus}_call_stats.out \
--coverage_file ${MuTect_dir}${Blood}${Nevus}coverage.wig.txt \
--vcf ${MuTect_dir}${Blood}${Nevus}.vcf \
--analysis_type MuTect --reference_sequence ${GATK_hg} --cosmic ${COSMIC} --dbsnp ${dbSNP} --intervals ${Intervals} \
--disable_auto_index_creation_and_locking_when_reading_rods \
--min_qscore 30 --max_alt_alleles_in_normal_count 0

The results got more somatic mutations than when I run exactly same command except that I dropped the last line. So I went through all IGVs, and found that some somatic mutations identified are obviously germline mutations!

I feel so anxious. I don't have an idea how to debug this. Please any one help me.

Also, though I set --max_alt_alleles_in_normal_count 0, MuTect is still picking up mutations. The one on upper section is for tumor, and lower section for normal.

[russhh]

You will inevitably get germline variants called as somatic hits, it's just a case of trying to limit how many come up. Do many of the variants come up at regions containing repetition or indels? You could reduce their influence by merging the tumour and normal data, running realignment, and then splitting them back up.

Conversely, if you look through the data yourself, do variants at TP53 / RB1 / KRAS (? not sure what the best candidate would be in your case) get identified by your workflow?

Answer 1

[haiying.kong]

Ok, I figured out. Those alternative alleles in normal sample in the locations that are identified as having somatic mutations are all having phred scores less than 30, which is another parameter I set when I ran MuTect. The alternative alleles are not considered when the software was deciding somatic mutations, but are all shown on IGV.

[igor]

MuTect will output all variants. You should filter for those that are flagged as PASS to see the ones it considers to be true somatic.

[scchess]

Yes. Apply FilterMutectCalls.

[igor]

Just to clarify, FilterMutectCalls was introduced in GATK4 (Mutect2). The original question used MuTect 1, which did not have filtering as a separate step.