Here is different calling result on specific tri-allelic site using GATK v3.6 and v3.7.
[ Command ]
java -XX:ParallelGCThreads=6 -Xmx30g -jar GenomeAnalysisTK-3.6 (3.7) /GenomeAnalysisTK.jar
--disable_auto_index_creation_and_locking_when_reading_rods
-T MuTect2
-R hg19.fa
-I:tumor tumor.bam
--normal_panel normal_pane.vcf
--dbsnp dbsnp_138.hg19.vcf
--cosmic CosmicCodingMuts.sorted.vcf
--bamOutput v36 (v37)
-o v36.vcf (v37.vcf)
-L chr7:55,084,725-55,277,031
[ MuTect2 in GATK v3.6 ]
chr7 55266417 . T A . panel_of_normals;triallelic_site ECNT=1;HCNT=6;MAX_ED=.;MIN_ED=.;NLOD=0.00;TLOD=6.49 GT:AD:AF:ALT_F1R2:ALT_F2R1:FOXOG:QSS:REF_F1R2:REF_F2R1 0/1:0,12:1.00:0:2:0.00:0,29:0:0
[ MuTect2 in GATK v3.7 ]
chr7 55266417 rs1140475 T C . panel_of_normals DB;ECNT=1;HCNT=55;MAX_ED=.;MIN_ED=.;TLOD=648.87 GT:AD:AF:ALT_F1R2:ALT_F2R1:FOXOG:QSS:REF_F1R2:REF_F2R1 0/1:0,340:1.00:162:177:0.478:0,9585:0:0
[ Variant in PON, COSMIC, and dbSNP ]
I think 'A' allele is somatic mutation, but the result of GATK v3.7 is 'C' might germline mutation.
I've searched GATK forum, and found the post related to change method calling on triallelic site.
(http://gatkforums.broadinstitute.org/gatk/discussion/8386/mutect2-and-alternate-allele-calls)
I want to know what is difference of calling algorithm between GATK v3.6 and v3.7 on tri-allelic site.
Always thank GATK team 