Missing data per site

Question

[anna]

Hi,

I want to calculate statistics of missing data per each site in my vcf file.

Using vcftools --missing-site gives wrong stats for several sites.

Is there is any other way to calculate it?

Thank you!

I have 36 samples and here is an example of the vcftools --missing-site output for four positions below

chr1    10616   rs376342519     CCGCCGTTGCAAAGGCGCGCCG  C       194.49  PASS    AC=3;AF=0.083;AN=36;DB;DP=132;ExcessHet=0.1902;FS=0;InbreedingCoeff=0.4307;MLEAC=5;MLEAF=0.139;MQ=38.12;MQRankSum=-0.842;QD=24.31;SOR=1.329;VQSLOD=-6.944;culprit=MQ    GT:AD:DP:GQ     0/0:2,0:2:6     ./.:0,0:0:.     ./.:0,0:0:.     ./.:0,0:0:.     0/0:5,0:5:15    0/0:2,0:2:3    ./.:0,0:0:.      1/1:0,3:3:10    ./.:0,0:0:.     ./.:0,0:0:.     0/0:9,0:9:24    0/0:11,0:11:27  ./.:0,0:0:.     ./.:0,0:0:.     0/0:9,0:9:15    0/0:19,0:19:54  0/0:4,0:4:12    0/0:9,0:9:2     ./.:0,0:0:.     0/0:9,0:9:9     0/0:8,0:8:18   0/0:8,0:8:24     0/0:6,0:6:15    ./.:0,0:0:.     ./.:0,0:0:.     0/0:2,0:2:3     0/0:6,0:6:6     ./.:0,0:0:.     ./.:0,0:0:.     ./.:0,0:0:.     0/0:8,0:8:24    ./.:0,0:0:.     0/1:3,2:5:76    ./.:0,0:0:.     ./.:0,0:0:.     ./.:2,0:2:.
chr2    120680534       rs74654922      A       ATGTGTGTG       12887.7 PASS    AC=7;AF=0.175;AN=40;DB;DP=751;ExcessHet=3.0103;FS=0;InbreedingCoeff=0.5714;MLEAC=12;MLEAF=0.3;MQ=58.52;NEGATIVE_TRAIN_SITE;QD=29.16;SOR=0.883;VQSLOD=-1.646;culprit=MQ  GT:AD:DP:GQ     1|1:0,10:12:30  0|0:0,1:17:48   .|.:0,2:16:.    0|0:0,1:21:60   0/0:0,1:6:16    ./.:38,0:38:.  0|0:0,0:24:72    ./.:21,0:21:.   0/0:0,0:13:39   .|.:0,1:23:.    0/0:0,2:11:48   0|0:0,1:22:63   .|.:0,0:18:.    ./.:20,0:20:.   ./.:15,0:15:.   1/0:0,12:30:99  0|0:0,0:12:36   1|1:0,19:20:57  .|.:0,1:25:.    ./.:22,0:22:.   0|0:0,0:19:57  0|0:0,0:13:39    .|.:0,1:16:.    0|0:0,0:26:78   ./.:26,0:26:.   0|0:0,0:29:87   0|0:0,1:29:84   1|1:0,19:21:57  .|.:0,0:23:.    0|0:0,2:23:63   ./.:14,0:14:.   0|0:0,0:20:60   ./.:11,0:11:.   0|0:0,0:15:45   ./.:32,0:32:.   ./.:35,0:35:.
chr22   50808269        .       AG      A       105.55  PASS    AC=2;AF=0.059;AN=34;DP=1217;ExcessHet=0.1296;FS=0;InbreedingCoeff=0.1633;MLEAC=7;MLEAF=0.206;MQ=36.17;NEGATIVE_TRAIN_SITE;QD=30.35;SOR=2.584;VQSLOD=-3.491;culprit=MQ   GT:AD:DP:GQ     ./.:22,0:22:.   0/0:38,0:38:0   ./.:24,0:24:.   ./.:20,0:20:.   ./.:26,0:26:.   ./.:43,0:43:.   ./.:26,0:26:.  ./.:19,0:19:.    0/0:30,0:30:0   ./.:26,0:26:.   ./.:27,0:27:.   0/0:43,0:43:0   0/0:22,0:22:0   ./.:31,0:31:.   ./.:28,0:28:.   0/0:34,0:34:0   ./.:26,0:26:.   1/1:0,2:7:7     0/0:51,0:51:0   0/0:33,0:33:0   ./.:21,0:21:.   ./.:9,0:9:.    0/0:22,0:22:0    0/0:28,0:28:0   0/0:41,0:41:0   0/0:61,0:61:0   ./.:54,0:54:.   0/0:43,0:43:0   ./.:36,0:36:.   0/0:52,0:52:0   ./.:17,0:17:.   0/0:51,0:51:0   ./.:19,0:19:.   0/0:63,0:63:0   ./.:34,0:34:.   0/0:43,0:43:0
chr18   67740193        .       C       CTTTTTTTTTTTTTT 8423.06 PASS    AC=9;AF=0.196;AN=46;DP=616;ExcessHet=0;FS=0;InbreedingCoeff=0.618;MLEAC=10;MLEAF=0.217;MQ=57.96;NEGATIVE_TRAIN_SITE;QD=26.16;SOR=0.956;VQSLOD=-2.121;culprit=MQ GT:AD:DP:GQ     .|.:0,4:15:.    .|.:0,5:15:.    1/1:0,2:5:42    .|.:0,2:14:.    1/1:0,5:6:0     0/0:29,0:29:0   0/0:15,0:15:6  0/0:0,1:6:14     0/0:13,0:13:0   .|.:0,4:12:.    0|0:0,6:26:60   .|.:0,3:11:.    .|.:0,1:6:.     0/0:0,1:4:20    1|1:0,20:25:60  0/0:18,0:18:0   .|.:0,4:15:.    .|.:0,4:18:.    0/0:29,0:29:12  0/0:0,5:15:90   0/0:0,2:6:18    0/0:0,1:6:38   .|.:0,4:15:.     0|0:0,1:17:46   0/0:0,3:9:44    .|.:0,6:12:.    0|0:0,1:18:51   0/0:21,0:21:6   .|.:0,8:16:.    .|.:0,5:17:.    .|.:0,4:14:.    1/1:0,1:1:0     0|0:0,2:27:75   1/0:0,1:5:38    0/0:24,0:24:15  0/0:16,0:16:9

And here is the output of vcftools --missing-site

CHR     POS     N_DATA  N_GENOTYPE_FILTERED     N_MISS  F_MISS  
chr1    10616   72      0       36      0.5  
chr2    120680534       66      0       26      0.393939  
chr22   50808269        72      0       38      0.527778  
chr18   67740193        59      0       13      0.220339

When in reality it should be this

chr1    10616   72      0       36      0.5  
chr2    120680534       72      0       32    0.44  
chr22   50808269        72      0       38      0.527778  
chr18   67740193        72      0       26      0.361

[Carlo Yague]

Why do you think it is wrong with vcftools --missing-site exactly ?

[anna]

Made edits in the question with examples

[Carlo Yague]

Thanks for elaborating. Looks like there is a bug, and the tool is counting only half of the missing allele for unphased genotypes (indicated with .|.). It is counting properly for the phased genotype (./.) however. I don't know if there is another tool to do this, but you could try to update vcftools in case the issue got fixed in a newer version. It should also be relatively easy to write a small script that gives you the proper answer.

[anna]

Yes, the thing that it does not give a right calculation even for unphased data, so DO NOT use vcftools for missing site calculations :)

I might consider awk instead

[Carlo Yague]

Crossposted on stackoverflow.

Answer 1

[anna]

This python code would work for this example:

import sys

infile = sys.argv[1]
try:
    ofile = sys.argv[2]
except IndexError:
    ofile = "/dev/stdout"

with open(infile, 'r') as fin:
    with open(ofile, 'w') as ofile:
        ofile.write("\t".join(["CHR", "POS", "N_DATA", "N_GENOTYPE_FILTERED", "N_MISS", "F_MISS\n"]))

        for line in fin:

            if line.startswith("#"): 
                continue

            line = line.strip().split()
            chrom = line[0]
            pos = line[1]
            missing = 0

            for smp in line[9:]:
                missing = missing + smp.split(":")[0].replace("|", "/").split("/").count(".")

            total = 2 * len(line[9:])

            oline = "\t".join([chrom, pos, str(total), "0", str(missing),  str(round(missing / total, 6)), "\n"])
            ofile.write(oline)

Assuming it's in a file called. e.g. missing_sites.py, it can be invoked by:

python missing_sites.py input.vcf 

or

python missing_sites.py input.vcf  outfile

Answer 2

[sbstevenlee]

Here's another Python solution using the pyvcf submodule from the fuc package I wrote:

>>> from fuc import pyvcf
>>> import pandas as pd
>>> 
>>> data = {
...     'CHROM': ['chr1', 'chr1', 'chr1'],
...     'POS': [100, 101, 102],
...     'ID': ['.', '.', '.'],
...     'REF': ['G', 'T', 'A'],
...     'ALT': ['A', 'C', 'C'],
...     'QUAL': ['.', '.', '.'],
...     'FILTER': ['.', '.', '.'],
...     'INFO': ['.', '.', '.'],
...     'FORMAT': ['GT', 'GT', 'GT'],
...     'A': ['0/0', './.', './.'],
...     'B': ['0/0', '1/1', '1/1'],
...     'C': ['0/1', './.', '0/0'],
... }
>>> vf = pyvcf.VcfFrame.from_dict([], data)
>>> # vf = pyvcf.VcfFrame.from_file('in.vcf')
>>> vf.df
  CHROM  POS ID REF ALT QUAL FILTER INFO FORMAT    A    B    C
0  chr1  100  .   G   A    .      .    .     GT  0/0  0/0  0/1
1  chr1  101  .   T   C    .      .    .     GT  ./.  1/1  ./.
2  chr1  102  .   A   C    .      .    .     GT  ./.  1/1  0/0
>>> def one_row(row):
...     m = row[9:].apply(pyvcf.gt_miss).mean()
...     s = pd.Series([row['CHROM'], row['POS'], m])
...     return s
... 
>>> df = vf.df.apply(one_row, axis=1)
>>> df.columns = ['CHROM', 'POS', 'Missing']
>>> df
  CHROM  POS   Missing
0  chr1  100  0.000000
1  chr1  101  0.666667
2  chr1  102  0.333333

Answer 3

[4galaxy77]

Almost certainly faster and easier to use plink2:

plink2 
    --vcf in.vcf \ 
    --missing \
    --out missing.out