Melt VCF files into a reasonable tab delimited format, one genotype per line

melt_vcf.py

#!/usr/bin/env python
""" Melt a VCF file into a tab delimited set of calls, one per line

VCF files have all the calls from different samples on one line.  This 
script reads vcf on stdin and writes all calls to stdout in tab delimited 
format with one call in one sample per line.  This makes it easy to find 
a given sample's genotype with, say, grep.
"""

import sys
import csv

out = csv.writer(sys.stdout, delimiter='\t')

fixed = None
samples = None
format = None

def parse_samples(line):
    toks = line[1:].split('\t')
    fixed, samples = toks[:9], toks[9:]
    return fixed, samples 
    
def parse_format(line):
    return line.split('\t')[8].split(':')

def parse_variant(line):
    return line.split('\t')[:8]

def parse_calls(line):
    for call in line.split('\t')[9:]:
        yield call.split(':')
    
for line in sys.stdin: 
    line = line.rstrip()
    if line.startswith('#'):
        if not line.startswith('##'):
            fixed, samples = parse_samples(line)
        
    else:
        if not format: 
            format = parse_format(line)
            out.writerow(['SAMPLE'] + format + fixed)
        
        variant = parse_variant(line)
        for sample, call in zip(samples, parse_calls(line)):
            while len(call) < len(format): 
                call.append('.')
            
            out.writerow([sample] + call + variant)
            

This was fantastically useful. Thanks!

It turned out not to work on the full length file. I think the problem had to do with line breaks; there were only line breaks between each variant in the vcf file, but the output contained lines that didn't start with reference IDs (i.e. chr1).