VCF combination for common variants

Question

[windsur]

Hello all!

I have several VCF files from the same patient (not identicals). The purpose is to combine the different VCF in one, but keeping only the common ones in at least 2 VCFs. I have tried with bcftools, and also with findCommonVariants of Rsubread library. But they give me only the commons in all, or I do not know how to obtain the intersections.

My VCF files are formed from Normal-Tumor pair fastq, here a little example (I will have n VCF files):

VCF1:

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  NORMAL  TUMOR
chr1    1013466 .   T   A   .   PASS    SOMATIC;QSS=49;TQSS=1;NT=ref;QSS_NT=49;TQSS_NT=1;SGT=TT->AT;DP=133;MQ=60.00;MQ0=0;ReadPosRankSum=-2.44;SNVSB=0.00;SomaticEVS=7.18   DP:FDP:SDP:SUBDP:AU:CU:GU:TU    32:0:0:0:0,0:0,0:0,0:32,32  100:0:0:0:5,5:0,0:0,0:95,96
chr1    1264801 .   A   G   .   PASS    SOMATIC;QSS=50;TQSS=1;NT=ref;QSS_NT=50;TQSS_NT=1;SGT=AA->AG;DP=132;MQ=60.00;MQ0=0;ReadPosRankSum=-2.09;SNVSB=0.00;SomaticEVS=10.57  DP:FDP:SDP:SUBDP:AU:CU:GU:TU    50:0:0:0:50,50:0,0:0,0:0,0  80:2:0:0:73,76:0,0:5,6:0,0
chr1    2312653 .   G   A   .   PASS    SOMATIC;QSS=16;TQSS=1;NT=ref;QSS_NT=16;TQSS_NT=1;SGT=GG->AG;DP=14;MQ=60.00;MQ0=0;ReadPosRankSum=-1.55;SNVSB=0.00;SomaticEVS=8.28    DP:FDP:SDP:SUBDP:AU:CU:GU:TU    10:0:0:0:0,0:0,0:10,10:0,0  4:0:0:0:2,2:0,0:2,2:0,0

VCF2:

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  26-M    7-M
chr1    1013466 .   A   G   .   PASS    AS_FilterStatus=SITE;AS_SB_TABLE=9,8|5,0;DP=23;ECNT=1;GERMQ=21;MBQ=33,30;MFRL=352,345;MMQ=23,40;MPOS=38;NALOD=0.778;NLOD=3.00;POPAF=6.00;TLOD=10.60 GT:AD:AF:DP:F1R2:F2R1:SB    0/0:10,0:0.083:10:7,0:3,0:2,8,0,0   0/1:7,5:0.428:12:3,5:4,0:7,0,5,0
chr1    1738127 .   C   T   .   PASS    AS_FilterStatus=SITE;AS_SB_TABLE=38,10|6,5;DP=61;ECNT=1;GERMQ=56;MBQ=35,20;MFRL=313,238;MMQ=60,45;MPOS=50;NALOD=1.15;NLOD=3.91;POPAF=6.00;TLOD=18.09    GT:AD:AF:DP:F1R2:F2R1:SB    0/0:14,0:0.067:14:6,0:8,0:12,2,0,0  0/1:34,11:0.194:45:17,6:17,4:26,8,6,5
chr1    2312653 .   G   A   .   PASS    AS_FilterStatus=SITE;AS_SB_TABLE=17,13|3,2;DP=35;ECNT=1;GERMQ=33;MBQ=20,20;MFRL=179,170;MMQ=58,60;MPOS=53;NALOD=0.997;NLOD=2.66;POPAF=6.00;TLOD=9.52    GT:AD:AF:DP:F1R2:F2R1:SB    0/0:13,0:0.092:13:4,0:8,0:7,6,0,0   0/1:17,5:0.212:22:9,4:7,1:10,7,3,2

VCF3:

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  N   T
chr1    1013466 .   T   A   .   PASS    SOMATIC;QSS=49;TQSS=1;NT=ref;QSS_NT=49;TQSS_NT=1;SGT=TT->AT;DP=133;MQ=60.00;MQ0=0;ReadPosRankSum=-2.44;SNVSB=0.00;SomaticEVS=7.18   DP:FDP:SDP:SUBDP:AU:CU:GU:TU    32:0:0:0:0,0:0,0:0,0:32,32  100:0:0:0:5,5:0,0:0,0:95,96
chr1    1738127 .   C   T   .   PASS    SOMATIC;QSS=50;TQSS=1;NT=ref;QSS_NT=50;TQSS_NT=1;SGT=AA->AG;DP=132;MQ=60.00;MQ0=0;ReadPosRankSum=-2.09;SNVSB=0.00;SomaticEVS=10.57  DP:FDP:SDP:SUBDP:AU:CU:GU:TU    50:0:0:0:50,50:0,0:0,0:0,0  80:2:0:0:73,76:0,0:5,6:0,0
chr1    2312847 .   C   T   .   PASS    SOMATIC;QSS=16;TQSS=1;NT=ref;QSS_NT=16;TQSS_NT=1;SGT=GG->AG;DP=14;MQ=60.00;MQ0=0;ReadPosRankSum=-1.55;SNVSB=0.00;SomaticEVS=8.28    DP:FDP:SDP:SUBDP:AU:CU:GU:TU    10:0:0:0:0,0:0,0:10,10:0,0  4:0:0:0:2,2:0,0:2,2:0,0

As the content of the last 4 columns are not the same in the other VCF files I think it can be done if the columns 1-7 from VCF1 are the same in at least one of the rest VCF, then copy columns 8-11 of the commons VCF to the VCF1.

VCFcombined: Output expected (It will be great if it is possible to add a column named FILE with the filenames presents):

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  NORMAL  TUMOR   FILTER  INFO    FORMAT  26-M    7-M N   T   FILE
    chr1    1013466 .   T   A   .   PASS    SOMATIC;QSS=49;TQSS=1;NT=ref;QSS_NT=49;TQSS_NT=1;SGT=TT->AT;DP=133;MQ=60.00;MQ0=0;ReadPosRankSum=-2.44;SNVSB=0.00;SomaticEVS=7.18   DP:FDP:SDP:SUBDP:AU:CU:GU:TU    32:0:0:0:0,0:0,0:0,0:32,32  100:0:0:0:5,5:0,0:0,0:95,96 AS_FilterStatus=SITE;AS_SB_TABLE=9,8|5,0;DP=23;ECNT=1;GERMQ=21;MBQ=33,30;MFRL=352,345;MMQ=23,40;MPOS=38;NALOD=0.778;NLOD=3.00;POPAF=6.00;TLOD=10.60 GT:AD:AF:DP:F1R2:F2R1:SB    0/0:10,0:0.083:10:7,0:3,0:2,8,0,0   0/1:7,5:0.428:12:3,5:4,0:7,0,5,0    VCF1,VCF2,VCF3
    chr1    2312653 .   G   A   .   PASS    SOMATIC;QSS=16;TQSS=1;NT=ref;QSS_NT=16;TQSS_NT=1;SGT=GG->AG;DP=14;MQ=60.00;MQ0=0;ReadPosRankSum=-1.55;SNVSB=0.00;SomaticEVS=8.28    DP:FDP:SDP:SUBDP:AU:CU:GU:TU    10:0:0:0:0,0:0,0:10,10:0,0  4:0:0:0:2,2:0,0:2,2:0,0 AS_FilterStatus=SITE;AS_SB_TABLE=17,13|3,2;DP=35;ECNT=1;GERMQ=33;MBQ=20,20;MFRL=179,170;MMQ=58,60;MPOS=53;NALOD=0.997;NLOD=2.66;POPAF=6.00;TLOD=9.52    GT:AD:AF:DP:F1R2:F2R1:SB    0/0:13,0:0.092:13:4,0:8,0:7,6,0,0   0/1:17,5:0.212:22:9,4:7,1:10,7,3,2  VCF1,VCF2
    chr1    1738127 .   C   T   .   PASS        .        .  AS_FilterStatus=SITE;AS_SB_TABLE=38,10|6,5;DP=61;ECNT=1;GERMQ=56;MBQ=35,20;MFRL=313,238;MMQ=60,45;MPOS=50;NALOD=1.15;NLOD=3.91;POPAF=6.00;TLOD=18.09    GT:AD:AF:DP:F1R2:F2R1:SB    0/0:14,0:0.067:14:6,0:8,0:12,2,0,0  0/1:34,11:0.194:45:17,6:17,4:26,8,6,5   VCF2,VCF3

Any help is more than welcome! Thanks!!

[g.papp-co]

Hi!

I'm sure R can do this as well, but pandas module has a merge function like:

df_merged = pd.merge(df_vcf1, df_vcf2, on=['Chrom', 'Pos', ...], how='inner')

You can do the same with the other combinations and at the end, concat them and drop duplicates. Maybe you can add extra column for the filenames before.

[Ram]

bcftools isec should be able to give you "present in n files" variants with the -n parameter

bcftools isec -n+2 -p output_dir *.vcf.gz

Ensure that your VCF is decomposed and normalized (left aligned, parsimonious representation) before you do this though, multi-allelic variants and non-normalized entries can mess up comparison.