Merge ca87919 into fdf8a3d

mutacc_auto/cli/export_command.py

@@ -0,0 +1,45 @@
+import click
+import logging
+from pathlib import Path
+
+import yaml
+
+from mutacc_auto.recipes.export_recipe import export_dataset
+
+
+LOG = logging.getLogger(__name__)
+
+def parse_path(ctx, param, value):
+
+    if value:
+        value = str(Path(str(value)).expanduser().absolute().resolve())
+    return value
+
+@click.command('export')
+@click.option('-C', '--mutacc-config',
+              type=click.Path(exists=True),
+              callback=parse_path,
+              help="configuration file used for mutacc")
+@click.option('-o', '--vcf-out',
+              type=click.Path(exists=False),
+              callback=parse_path,
+              help="Path to created vcf-file")
+@click.option('-b', '--background',
+              type=click.Path(exists=True),
+              help="yaml file with genomic backgrounds for each sample in trio")
+@click.pass_context
+def export_command(ctx, mutacc_config, vcf_out, background):
+
+    mutacc_config = mutacc_config or ctx.obj['mutacc_config']
+    mutacc_binary = ctx.obj.get('mutacc_binary')
+
+    with open(background, 'r') as background_handle:
+        background_datasets = yaml.load(background_handle)
+
+    files = export_dataset(mutacc_config=mutacc_config,
+                           background=background_datasets,
+                           mutacc_binary=mutacc_binary,
+                           case_query='{}',
+                           merged_vcf_path=vcf_out)
+
+    LOG.debug("files created: {}".format(files))

mutacc_auto/cli/root.py

@@ -7,6 +7,7 @@
 from mutacc_auto import __version__
 from .extract_command import extract_command
 from .import_command import import_command
+from .export_command import export_command
 
 LOG_LEVELS = ['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL']
 LOG = logging.getLogger(__name__)
@@ -36,3 +37,4 @@ def cli(ctx, loglevel, config_file):
 
 cli.add_command(extract_command)
 cli.add_command(import_command)
+cli.add_command(export_command)

mutacc_auto/commands/command.py

@@ -73,7 +73,7 @@ def check_output(self):
 
         command_stdout = command_stdout.decode('utf-8')
 
-        LOG.debug("Executed:".format(str(self)))
+        LOG.debug("Executed: {}".format(str(self)))
 
         return command_stdout
 
@@ -96,6 +96,6 @@ def call(self):
 
             raise
 
-        LOG.debug("Executed:".format(str(self)))
+        LOG.debug("Executed: {}".format(str(self)))
 
         return result

mutacc_auto/commands/mutacc_command.py

@@ -37,3 +37,41 @@ def __init__(self, config_file, extracted_case_file, mutacc_binary=None):
         self.add_subcommand('db')
         self.add_subcommand('import')
         self.add_argument(str(extracted_case_file))
+
+class MutaccExport(MutaccCommand):
+
+    def __init__(self, config_file, mutacc_binary=None, case_query=None, variant_query=None,
+                 proband=False, member='affected', sample_name=None):
+
+        super(MutaccExport, self).__init__(config_file, mutacc_binary=mutacc_binary)
+
+        self.add_subcommand('db')
+        self.add_subcommand('export')
+
+        if variant_query is not None:
+            self.add_option('variant-query', variant_query)
+        if case_query is not None:
+            self.add_option('case-query', case_query)
+        if sample_name is not None:
+            self.add_option('sample-name', sample_name)
+        if proband:
+            self.add_option('proband')
+
+        self.add_option('member', member)
+        self.add_option('json-out')
+
+class MutaccSynthesize(MutaccCommand):
+
+    def __init__(self, config_file, fastq1, bam_file, query_file, fastq2=None,
+                 mutacc_binary=None):
+
+        super(MutaccSynthesize, self).__init__(config_file, mutacc_binary=mutacc_binary)
+
+        self.add_subcommand('synthesize')
+
+        self.add_option('background-bam', bam_file)
+        self.add_option('background-fastq', fastq1)
+        if fastq2 is not None:
+            self.add_option('background-fastq2', fastq2)
+        self.add_option('query', query_file)
+        self.add_option('json-out')

mutacc_auto/commands/vcf_command.py

@@ -0,0 +1,37 @@
+from .command import Command as BaseCommand
+
+BGZIP_COMMAND = 'bgzip'
+TABIX_COMMAND = 'tabix'
+BCFTOOLS_MERGE_COMMAND = 'bcftools'
+
+class BgzipCommand(BaseCommand):
+
+    def __init__(self, vcf_file):
+
+        super(BgzipCommand, self).__init__(BGZIP_COMMAND)
+
+        self.add_argument(vcf_file)
+
+class TabixCommand(BaseCommand):
+
+    def __init__(self, vcf_file):
+
+        super(TabixCommand, self).__init__(TABIX_COMMAND)
+
+        self.add_argument(vcf_file)
+
+class BcftoolsMergeCommand(BaseCommand):
+
+    def __init__(self, vcf_files, out_vcf=None):
+
+        super(BcftoolsMergeCommand, self).__init__(BCFTOOLS_MERGE_COMMAND)
+
+        self.add_subcommand('merge')
+
+        if out_vcf:
+            self.add_option('output', out_vcf)
+
+        self.add_option('output-type', 'z')
+
+        for vcf_file in vcf_files:
+            self.add_argument(vcf_file)

mutacc_auto/recipes/export_recipe.py

@@ -0,0 +1,226 @@
+"""
+    Functions to export synthetic dataset
+"""
+
+import logging
+import json
+import os
+from pathlib import Path
+
+from mutacc_auto.commands.mutacc_command import (MutaccExport, MutaccSynthesize)
+from mutacc_auto.commands.vcf_command import (BgzipCommand, TabixCommand, BcftoolsMergeCommand)
+
+LOG = logging.getLogger(__name__)
+
+
+def run_mutacc_export_command(mutacc_config, mutacc_binary=None, case_query=None,
+                              variant_query=None, proband=False, member='affected',
+                              sample_name=None):
+
+    """
+        Runs the 'mutacc db export' command
+
+        Args:
+            mutacc_config (Path): Path to mutacc config file
+            mutacc_binary (Path): path to mutacc binary
+            case_query (string): json string with query against case collection
+            variant_query (string): json string with query against variant collection
+            proband (bool): True if sample is proband, False if not
+            member (string): affected|father|mother|child
+            sample_name (string): name of created sample
+
+        returns:
+            export_out (string): stdout from mutacc command
+
+    """
+
+    mutacc_export_command = MutaccExport(config_file=mutacc_config,
+                                         mutacc_binary=mutacc_binary,
+                                         case_query=case_query,
+                                         variant_query=variant_query,
+                                         proband=proband,
+                                         member=member,
+                                         sample_name=sample_name)
+    export_out = mutacc_export_command.check_output()
+    export_out = json.loads(export_out)
+    return export_out
+
+def export_trio(mutacc_config, mutacc_binary=None, case_query=None, variant_query=None):
+
+    """
+        Exports trio from mutaccDB
+
+        Args:
+            mutacc_config (Path): Path to mutacc config file
+            mutacc_binary (Path): path to mutacc binary
+            case_query (string): json string with query against case collection
+            variant_query (string): json string with query against variant collection
+
+        Returns:
+            mutacc_files (dict): files created by mutacc for each sample
+
+    """
+
+    members = ('child', 'father', 'mother')
+    mutacc_files = {}
+    for member in members:
+
+        member_files = run_mutacc_export_command(mutacc_config=mutacc_config,
+                                                 mutacc_binary=mutacc_binary,
+                                                 case_query=case_query,
+                                                 variant_query=variant_query,
+                                                 proband=True if member == 'child' else False,
+                                                 member=member,
+                                                 sample_name=member)
+
+        mutacc_files[member] = member_files
+
+    return mutacc_files
+
+def bgzip_vcf_file(vcf_file):
+
+    """
+        Uses command line tool 'bgzip' to compress vcf
+
+        Args:
+            vcf_file (Path): path to vcf
+        Returns:
+            vcf_gz_file (Path): path to compressed vcf
+    """
+
+    zip_command = BgzipCommand(vcf_file)
+    zip_command.call()
+    return f"{vcf_file}.gz"
+
+def index_vcf_file(vcf_file):
+
+    """
+        Uses command line tool 'tabix' to index vcf
+
+        Args:
+            vcf_file (Path): path to vcf
+        Returns:
+            vcf_index_file (Path): path to indexed vcf
+
+    """
+
+    index_command = TabixCommand(vcf_file)
+    index_command.call()
+    return f"{vcf_file}.tbi"
+
+def merge_vcf_files(vcf_files, out_file=None):
+
+    """
+        Merges vcf files using 'bcftools merge'
+
+        Args:
+            vcf_files (list): list of paths to vcf files
+
+        Returns:
+            out_file (Path): path to merged vcf file
+    """
+
+    merge_command = BcftoolsMergeCommand(vcf_files, out_vcf=out_file)
+    merge_command.call()
+    return out_file
+
+def synthesize_dataset(sample, mutacc_binary=None, mutacc_config=None):
+
+    """
+        Uses 'mutacc synthesize' to make synthetic dataset
+
+        Args:
+            sample (dict): Dictionary with fastq files, bam, and query file for sample
+            mutacc_config (Path): Path to mutacc config file
+            mutacc_binary (Path): path to mutacc binary
+
+        Returns:
+            dataset (list): list of fastq files created
+    """
+
+    synthesize_command = MutaccSynthesize(config_file=mutacc_config,
+                                          mutacc_binary=mutacc_binary,
+                                          fastq1=sample['fastq1'],
+                                          fastq2=sample['fastq2'],
+                                          bam_file=sample['bam'],
+                                          query_file=sample['query'])
+
+    dataset = synthesize_command.check_output()
+    dataset = json.loads(dataset)['fastq_files']
+    return dataset
+
+
+def synthesize_trio(mutacc_config, samples, mutacc_binary=None):
+    """
+        Synthesizes a trio
+
+        Args:
+            mutacc_config (Path): Path to mutacc config file
+            mutacc_binary (Path): path to mutacc binary
+            samples (dict(dict)): fastq files, bam, query for each sample in trio
+
+        Returns:
+            datasets (dict): fastq files for each sample
+
+    """
+    datasets = {}
+    for member in samples.keys():
+        dataset = synthesize_dataset(mutacc_config=mutacc_config,
+                                     mutacc_binary=mutacc_binary,
+                                     sample=samples[member])
+        datasets[member] = dataset
+
+    return datasets
+
+
+def export_dataset(mutacc_config, background=None, mutacc_binary=None, case_query=None,
+                   variant_query=None, merged_vcf_path=None):
+
+    """
+        Export a synthetic trio
+
+        Args:
+            mutacc_config (Path): Path to mutacc config file
+            mutacc_binary (Path): path to mutacc binary
+            background (dict): dictionary with background files to be used for each sample
+            case_query (string): json string with query against case collection
+            variant_query (string): json string with query against variant collection
+            merged_vcf_path (Path): path where to create vcf file
+
+        Returns:
+            datasets (dict): fastq files for each sample
+
+    """
+
+    files = export_trio(mutacc_config=mutacc_config,
+                        mutacc_binary=mutacc_binary,
+                        case_query=case_query,
+                        variant_query=variant_query)
+
+    vcf_files = [files[member]['vcf_file'] for member in files.keys()]
+    zipped_vcf_files = [bgzip_vcf_file(vcf_file) for vcf_file in vcf_files]
+    indexed_vcf_files = [index_vcf_file(vcf_file) for vcf_file in zipped_vcf_files]
+
+    vcf_path = Path.cwd().joinpath('merged_mutacc_set.vcf.gz')
+    if merged_vcf_path:
+        vcf_path = merged_vcf_path
+
+    vcf_path = merge_vcf_files(vcf_files=zipped_vcf_files, out_file=vcf_path)
+
+    # Remove individual vcf files and their indices
+    for zipped, indexed in zip(zipped_vcf_files, indexed_vcf_files):
+        os.remove(zipped)
+        os.remove(indexed)
+
+    print("FILES: ", files)
+    print("BACKGROUNDS: ", background)
+    samples = {member: {'fastq1': background[member]['fastq1'],
+                        'fastq2': background[member]['fastq2'],
+                        'bam': background[member]['bam'],
+                        'query': files[member]['query_file']} for member in files.keys()}
+
+    datasets = synthesize_trio(mutacc_config=mutacc_config,
+                               samples=samples,
+                               mutacc_binary=mutacc_binary)
+
+    return datasets