This site is now readonly. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!
Metochontrial WDL flow allele fraction(AF) is inconcistent with AD?
I had some metochontrial samples, and do GATK meto flow as GATK WDL.
then I find the allele fraction of AF is inconcistent with real alt_fraction,that is AD[2]/DP：
for example,AF of this record is 8.597*10^4, but real alt_fraction is 78/36748~~0.002,which should be not filtered by low_allele_frac.chrM 149 . T C . low_allele_frac CONTQ=15;DP=37739;ECNT=29;MBQ=20,20;MFRL=189,190;MMQ=60,60;MPOS=28;OCM=0;POPAF=2.40;SEQQ=93;STRANDQ=2;TLOD=12.96 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:36670,78:8.597e04:36748:16938,37:19352,39:21680,14990,47,31
then I check other records in the VCF, then I find their AFs are always inconcistent with AD/DP，like below:
this is visible format:
this is real vcf record of above:chrM 1 . G C . contamination;strand_bias CONTQ=2;DP=15294;ECNT=29;MBQ=20,20;MFRL=169,157;MMQ=60,60;MPOS=12;OCM=0;POPAF=2.40;SEQQ=93;STRANDQ=1;TLOD=26.51 GT:AD:AF:DP:F1R2:F2R1:PGT:PID:PS:SB 01:15268,26:1.611e03:15294:7295,11:7350,11:01:8568_T_C:8568:10035,5233,18,8 chrM 7 . A G . contamination;strand_bias CONTQ=1;DP=16204;ECNT=29;MBQ=20,20;MFRL=169,157;MMQ=60,60;MPOS=5;OCM=0;POPAF=2.40;SEQQ=93;STRANDQ=1;TLOD=25.52 GT:AD:AF:DP:F1R2:F2R1:PGT:PID:PS:SB 01:15848,26:1.541e03:15874:7166,13:7088,13:01:8568_T_C:8568:10437,5411,18,8 chrM 14 . T C . strand_bias CONTQ=29;DP=3333;ECNT=24;MBQ=20,29;MFRL=158,194;MMQ=60,60;MPOS=9;OCM=0;POPAF=2.40;SEQQ=52;STRANDQ=1;TLOD=5.14 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:2939,12:3.283e03:2951:1467,3:1438,9:1844,1095,5,7 chrM 16 . A G . strand_bias CONTQ=3;DP=3716;ECNT=24;MBQ=20,37;MFRL=159,245;MMQ=60,60;MPOS=8;OCM=0;POPAF=2.40;SEQQ=23;STRANDQ=1;TLOD=2.05 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:3353,6:1.818e03:3359:1667,3:1640,3:2113,1240,4,2 chrM 30 . T C . pon CONTQ=64;DP=17980;ECNT=29;MBQ=20,20;MFRL=165,159;MMQ=60,60;MPOS=30;OCM=0;POPAF=2.40;SEQQ=93;STRANDQ=5;TLOD=18.04 GT:AD:AF:DP:F1R2:F2R1:SB 0/1:17311,65:2.009e03:17376:8674,38:8437,26:10619,6692,42,23
so what's happened? I am appreciate for any help.
my mutect2 parameters are as below, I disable downsampling, I don't know if this introduce the matter?/usr/local/gatk4.1.4.0/gatk javaoptions "Xmx5g" Mutect2 R /database/GATK_bundle/chrM/references_hg38_v0_chrM_Homo_sapiens_assembly38.chrM.fasta I bwa/M627.rmdup_ref.bam O mutect2/M627.m2_ref.vcf bamoutput mutect2/M627.m2_ref.bam annotation StrandBiasBySample mitochondriamode independentmates **_maxreadsperalignmentstart 0_** maxmnpdistance 0
Best Answer

bhanuGandham Cambridge MA admin
Hi @ahda
You should use AD/DP because AF is the maxlikelihood estimate of the allele fraction given that the variant exists. That is, its purpose is to characterize variants, not to be used for filtering.
Answers
Hi @ahda
Mutect2 emits a probabilistic estimate of AF; not AD/DP.
More info on this can be found here: https://www.biorxiv.org/content/10.1101/861054v1
@bhanuGandham Thanks for reply.
It's a little hard for me to understand.
I found that the trend between AF and AD/DP is almost same,just a little difference in low alt fraction variants. Do the AF and AD/DP have some relationship?
I want to detect (alt fraction)>0.001 metochontrial SNP/Indels.
Should I use AF directly by AF>0.001? or AD/DP>0.001?
Can you kindly give me some suggestion?
Hi @ahda
You should use AD/DP because AF is the maxlikelihood estimate of the allele fraction given that the variant exists. That is, its purpose is to characterize variants, not to be used for filtering.
Thank you. I got it!