Extract variant positions from VCF dependent on contents of other columns

Question

I have a vcf file, I am trying to extract the information from these columns:

#CHROM  POS   REF     ALT

However I would like to extract these only if the SAMPLE-1 column contains the string DeNovo (Not DeNovoSV) and that SAMPLE-1, SAMPLE-2, and SAMPLE-3 all contain PASS.

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  SAMPLE-1  SAMPLE-2  SAMPLE-3
chr1    10230   .       AC      A       186.90  .       AC=4;AF=0.667;AN=6;DP=77;FS=0.000;MQ=26.38;MQRankSum=0.436;QD=3.89;ReadPosRankSum=0.000;SOR=1.162       GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP
:DN 0/1:6,12:0.667:18:32:PASS:2,3:4,9:69,0,30:3.9669e+01,2.7888e-03,3.2724e+01:295,0,215:Inherited  0/1:5,15:0.750:20:11:PASS:3,6:2,9:60,0,8:3.0340e+01,3.6694e-01,1.0964e+01:172,0,137:.   1/1:0,10
:1.000:10:26:PASS:0,2:0,8:93,29,0:6.1212e+01,2.6342e+01,1.0101e-02:378,0,183:.
chr1    61871   .       C       CT      60.27   .       AC=3;AF=0.500;AN=6;DP=29;FS=11.290;MQ=33.00;MQRankSum=-0.423;QD=2.51;ReadPosRankSum=0.705;SOR=0.478     GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP:DPL:DN:DQ  0/0:5,0:0.000:5:15:PASS:.:.:0,15,182:.:0,7,93:0,15,100:DeNovo:2.9227e-07        1/1:0,2:1.000:2:5:PASS:0,1:0,1:42,6,0:2.4787e+01,4.7870e+00,1.7754e+00:29,0,9:40,6,0:.  0/1:15,7:0.318:22:26:PASS:6,3:9,4:43,0,41:2.6538e+01,9.8206e-03,4.4010e+01:65,0,250:74,0,234:.
chr1    66369   .       TA      T       116.77  .       AC=2;AF=0.500;AN=4;DP=56;FS=10.138;MQ=173.59;MQRankSum=1.468;QD=4.32;ReadPosRankSum=0.929;SOR=0.367     GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP    ./.:11,5:0.312:16:0:LowGQ:.:.   0/1:8,4:0.333:12:40:PASS:3,2:5,2:71,0,43:4.1762e+01,4.0824e-04,4.5625e+01:71,0,43       0/1:8,7:0.467:15:45:PASS:4,4:4,3:77,0,47:4.7400e+01,1.2244e-04,5.0000e+01:77,0,47
chr1    934273  .       G       C       8.67    .       AC=1;AF=0.167;AN=6;DP=26;FS=0.000;MQ=19.17;MQRankSum=-1.179;QD=0.96;ReadPosRankSum=1.666;SOR=0.223      GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP
:DPL:DN:DQ  0/1:7,2:0.222:9:11:PASS:5,1:2,1:45,0,32:1.0868e+01,3.7242e-01,3.5372e+01:45,0,60:45,0,32:DeNovoSV:4.3945e-09    0/0:6,3:0.333:9:0:LowGQ:.:.:0,0,140:.:55,0,191:46,0,186:.       0/0:8,0:
0.000:8:23:PASS:.:.:0,23,190:.:0,25,195:0,23,190:.
chr1    934274  .       G       C       8.68    .       AC=1;AF=0.167;AN=6;DP=26;FS=0.000;MQ=19.17;MQRankSum=-1.179;QD=0.96;ReadPosRankSum=1.666;SOR=0.223      GT:AD:AF:DP:GQ:FT:F1R2:F2R1:PL:GP:PP
:DPL:DN:DQ  0/1:7,2:0.222:9:11:PASS:5,1:2,1:45,0,32:1.0868e+01,3.7242e-01,3.5372e+01:45,0,60:45,0,32:DeNovoSV:4.3945e-09    0/0:6,3:0.333:9:0:PASS:.:.:0,0,140:.:55,0,191:46,0,186:.       0/0:8,0:
0.000:8:23:PASS:.:.:0,23,190:.:0,25,195:0,23,190:.

I have tried using bcftools, see below.

bcftools query -f '%CHROM %POS %REF %ALT\n' file.vcf | head -3
chr1 10230 AC A
chr1 61871 C CT
chr1 66369 TA T

Is there a way to use bcftools, or combine it with awk in order to get the output I am looking for in the vcf file format?

Many thanks

Answer 1

FWIW, I'm not familiar with bcftools but if the intention is to pipe the bcftools output into awk then we could do the whole thing in awk ...

Assumptions:

• none of the fields (in the file) contain white space
• always going to pull data for columns #CHROM / POS / REF / ALT
• these 4 columns are always going to be in the same fields/positions (#CHROM = $1 / POS = $2 / REF = $4 / ALT = $5)
• DeNovo always shows up between a pair of colons (ie, we're looking for :DeNovo: in field $10)
• the three test fields always show up in the same fields/positions (SAMPLE_1 = $10 / SAMPLE_2 = $11 / SAMPLE_3 = $12)
• PASS always shows up between a pair of colons (ie, we're lookinf for :PASS: in fields $10 / $11 / $12)

One awk idea:

awk '
$10 ~ ":DeNovo:" &&
$10 ~ ":PASS:"   &&
$11 ~ ":PASS:"   &&
$12 ~ ":PASS:"      { print $1,$2,$4,$5 }
' file.vcf

This generates:

chr1 61871 C CT

NOTE: The obvious (?) downside to this approach is that we've hardcoded the column references; we could certainly modify the awk code to provide a more dynamic interface (replicate bcftools functionality?) but I'm not sure the added complexity is worth the effort, ymmv.

---

Assuming OP needs the functionality of bcftools (eg, designate a variable set of columns), one idea would be to modify the current bcftools call to include the SAMPLE_X columns and then pipe that output to awk; something like:

bcftools query -f '%CHROM %POS %REF %ALT %SAMPLE_1 %SAMPLE_2 %SAMPLE_3\n' file.vcf | awk '
$5 ~ ":DeNovo:" &&
$5 ~ ":PASS:"   &&
$6 ~ ":PASS:"   &&
$7 ~ ":PASS:"      { print $1,$2,$3,$4 }'

NOTES:

• further improvements would include parameterizing the DeNovo and PASS strings
• going back to the idea of eliminating bcftools and doing the whole thing in awk ... would likely need two column lists (columns to display, columns to test) along with a list of the string(s) to compare each of the 'test' columns against ... doable but now we're getting into a fairly substantial coding effort (vs what we've done so far)