def combine_vcf_by_gatk(vcf_file_list, output_vcf, loadJava7=False):
"""
Only for hg19 now.
Run GATK to combine VCF files in vcf_file_list. The samples are REQURIED to be different in all
files in the vcf_file_list
"""
if loadJava7:
load_java7 = LOAD_JAVA7_COMMAND
else:
load_java7 = ''
gatk_command = '''
%(load_java7)s
java -Xmx4g -jar %(GATKCOMMAND)s \
-R %(REF)s \
-T CombineVariants \
%(vcf_file_statement)s \
-o %(output_vcf)s
''' % {
'GATKCOMMAND': GATK_JAR,
'REF': B37_REF,
'output_vcf': output_vcf,
'load_java7': load_java7,
'vcf_file_statement': create_gatk_vcf_list_statements(vcf_file_list)
}
run_subprocess_cmd(gatk_command)
return 'Done'
def fix_vcf_header_for_virmid(virmid_vcf):
"""
Remove the illegal space in the header line in the VCF file produced by VIRMID.
The original virmid_vcf will be changed in place.
"""
virmid_vcf = os.path.expanduser(virmid_vcf)
cmd = "sed -i 's/ID=gt,Number=A,Type=String, Description=/ID=gt,Number=A,Type=String,Description=/' %s" % virmid_vcf
run_subprocess_cmd(cmd)
def create_vcf_index(vcf_file):
"""
Uses igvtools to create an index file for vcf_file. Relies on $SACGF environmental viable.
igv.log will be removed """
cmds = '''
{IGVTOOLS_CMD} index {VCF_FILE}
rm -f igv.log
'''.format(IGVTOOLS_CMD=IGVTOOLS_CMD, VCF_FILE=vcf_file)
run_subprocess_cmd(cmds)
def validate_vcf_spec_by_gatk(infile):
"""To perform only VCF format tests, by GATK"""
gatk_command = '''
java -Xmx4g -jar %(GATKCOMMAND)s \
-R %(REF)s \
-T ValidateVariants \
--validationTypeToExclude ALL \
--variant %(infile)s
''' % {
'GATKCOMMAND': GATK_JAR,
'REF': B37_REF,
'infile': infile
}
run_subprocess_cmd(gatk_command) # GATK writes stdout to stderr
def select_variants_by_gatk(invcf, outvcf, is_freebayes_somatic_vcf=False, sample_list=None,
intervals=None, interval_padding=False, exclude_sample=False, hugeVCF=False, num_threads=None):
"""
Run GATK -T SelectVariants, specifing the sample_list and/or intervals.
When is_freebayes_somatic_vcf is False (the default), it uses --removeUnusedAlternates and --excludeNonVariants.
intervals is a file path (e.g. A BED file), a single interval string ('1:100-200'),
a single point position ('14:45645954') or a list of string like ['1:100-200', '14:45645954']
outvcf must be a vcf.gz file.
Only for hg19 now.
"""
#TODO remove the tmp_vcf_out when GATK fix the * non-variant issue
tmp_vcf_out = random_file_at_same_dir(outvcf, prefix='before_removing_star', extension='vcf.gz')
sample_statement = ''
if sample_list:
if exclude_sample:
for sample in sample_list:
sample_statement += '--exclude_sample_name %s ' % sample
else:
for sample in sample_list:
sample_statement += '--sample_name %s ' % sample
interval_statements = ''
if intervals:
if type(intervals) is list:
for interval in intervals:
interval_statements += ' --intervals %s ' % interval
else: # it is a file path or a single interval string
interval_statements += ' --intervals %s ' % intervals
if interval_padding:
interval_statements += ' --interval_padding %s ' % interval_padding
if hugeVCF:
java_Xmx = '-Xmx14g'
else:
java_Xmx = '-Xmx4g'
if num_threads:
nt_statement = '--num_threads %s' % num_threads
else:
nt_statement = ''
if is_freebayes_somatic_vcf:
remove_unused_alternates_option = ''
else:
remove_unused_alternates_option = '--removeUnusedAlternates --excludeNonVariants'
# GATK SelectVariants does normalizing selected variants
gatk_command = '''
java %(java_Xmx)s -jar %(GATKCOMMAND)s \
-R %(REF)s \
--variant %(invcf)s \
-o %(output_vcf)s \
%(sample_statement)s %(interval_statements)s \
%(nt_statement)s \
-T SelectVariants \
%(remove_unused_alternates_option)s
''' % {
'GATKCOMMAND': GATK_JAR,
'java_Xmx': java_Xmx,
'REF': B37_REF,
'invcf': invcf,
'output_vcf': tmp_vcf_out,
'sample_statement': sample_statement,
'nt_statement': nt_statement,
'interval_statements': interval_statements,
'remove_unused_alternates_option': remove_unused_alternates_option
}
run_subprocess_cmd(gatk_command)
#TODO remove the tmp_vcf_out when GATK fix the * non-variant issue
remove_star_cmd = '''
{bcftools} view -O z --exclude "ALT='*'" {invcf} > {outvcf}
{bcftools} index -t {outvcf}
'''.format(bcftools=BCFTOOLS, invcf=tmp_vcf_out, outvcf=outvcf)
run_subprocess_cmd(remove_star_cmd)
remove_vcf_file(tmp_vcf_out)
return 'Done'