default='local',
help='',
choices=('local', 'drmaa:ge', 'ge', 'slurm'))
p.add_argument('-q',
'--queue',
help='Submit to this queue of the DRM supports it')
args = p.parse_args()
cosmos = Cosmos(
'sqlite:///%s/sqlite.db' % os.path.dirname(os.path.abspath(__file__)),
# example of how to change arguments if you're NOT using default_drm='local'
get_submit_args=partial(default_get_submit_args, parallel_env='smp'),
default_drm=args.drm,
default_queue=args.queue)
cosmos.initdb()
sp.check_call('mkdir -p analysis_output/ex2', shell=True)
os.chdir('analysis_output/ex2')
workflow = cosmos.start('Example2', restart=True, skip_confirm=True)
recipe(workflow)
workflow.make_output_dirs()
workflow.run(max_cores=10)
# Noting here that if you wanted to look at the outputs of any Tasks to decide how to generate the rest of a DAG
# you can do so here, proceed to add more tasks via workflow.add_task(), and then call workflow.run() again.
# Yes, it does require running all Tasks in the dag to get the outputs of any Task, and we hope to address
# that limitation at some point in the future.
import subprocess as sp
import os
import sys
from cosmos.api import Cosmos
cosmos = Cosmos('sqlite:///%s/sqlite.db' % os.path.dirname(os.path.abspath(__file__)),
default_drm='local')
cosmos.initdb()
sp.check_call('mkdir -p analysis_output/ex1', shell=True)
os.chdir('analysis_output/ex1')
workflow = cosmos.start('Example1', restart=True, skip_confirm=True)
def say(text, out_file):
return r"""
echo "{text}" > {out_file}
""".format(text=text, out_file=out_file)
t = workflow.add_task(func=say,
params=dict(text='Hello World', out_file='out.txt',),
uid='my_task', time_req=None, core_req=1, mem_req=1024)
print('task.params', t.params)
print('task.input_map', t.input_map)
print('task.output_map', t.output_map)
print('task.core_req', t.core_req)
print('task.time_req', t.time_req)
print('task.drm', t.drm)
print('task.uid', t.uid)
class Pipeline(object):
def __init__(self, config, drm, restart, skip_confirm):
self.config = config
self.cosmos = Cosmos(database_url='sqlite:///{}'.format(
self.config.db),
get_submit_args=default_get_submit_args,
default_drm=drm)
self.cosmos.initdb()
primary_logfile = os.path.join(
self.config.rootdir,
'{}.log'.format(self.config.project_name),
)
self.workflow = self.cosmos.start(
self.config.project_name,
primary_log_path=primary_logfile,
restart=restart,
skip_confirm=skip_confirm,
)
self.setup_pipeline()
def setup_pipeline(self):
self.construct_pipeline()
self.workflow.make_output_dirs()
def run(self, task_flush):
# put set_successful to False if you intend to add more tasks to the
# pipeline later
custom_log_dir = lambda task: os.path.join(self.config.rootdir, 'logs',
task.stage.name, task.uid)
self.workflow.run(set_successful=False,
log_out_dir_func=custom_log_dir,
db_task_flush=task_flush)
def construct_pipeline(self):
# 1. remove unused alternates
remove_ac_0_tasks = self.create_remove_ac_0_tasks(1)
# 2. calculate sample missingness (counting phase)
count_sample_missingness_tasks = self.create_count_sample_missingness_tasks(
remove_ac_0_tasks, 2)
# 2.1 calculate sample missingness (merge and calculation phase)
calculate_sample_missingness_task = self.create_calculate_sample_missingness_task(
count_sample_missingness_tasks, 2.1)
# 3. denormalize, decompose, and uniq
dnu_tasks = self.create_decompose_normalize_unique_tasks(
remove_ac_0_tasks, 3)
# 4. remove symbolic alleles
rsa_tasks = self.create_remove_symbolic_deletion_tasks(dnu_tasks, 4)
# 5. filter missingness
filter_variant_missingness_tasks = self.create_filter_variant_missingness_tasks(
rsa_tasks, 5)
# 6. annotate allele balances
allele_balance_annotation_tasks = self.create_allele_balance_annotation_tasks(
filter_variant_missingness_tasks, 6)
# 7. annotate with 1000G
annotate_1000G_tasks = self.create_1000G_annotation_tasks(
allele_balance_annotation_tasks, 7)
# 8. annotate with ExAC
annotate_ExAC_tasks = self.create_ExAC_annotation_tasks(
annotate_1000G_tasks, 8)
# 9. VEP annotation
annotate_vep_cadd_tasks = self.create_vep_cadd_annotation_tasks(
annotate_ExAC_tasks, 9)
# 10. VCF concatenation
concatenated_vcfs = self.create_concatenate_vcfs_task(
annotate_vep_cadd_tasks, 10)
# 11. bcftools stats
bcftools_stats_tasks = self.create_bcftools_stats_tasks(
annotate_ExAC_tasks, 11)
# 11.1 Merge & Plot bcftools stats
bcftools_stats_summary_task = self.create_bcftools_stats_summary_task(
bcftools_stats_tasks, 11.1)
# 12. GATK VariantEval
variant_eval_tasks = self.create_variant_eval_tasks(
annotate_ExAC_tasks, 12)
# 12.1. Merge & Plot GATK VariantEval Stats
variant_eval_summary_task = self.create_variant_eval_summary_task(
variant_eval_tasks, 12.1)
def create_bcftools_stats_summary_task(self, parent_tasks, step_number):
stage = self._construct_task_name('bcftools-stats-summary',
step_number)
output_dir = os.path.join(self.config.rootdir, stage)
prior_stage_name = parent_tasks[0].stage.name
input_dir = os.path.join(self.config.rootdir, prior_stage_name)
lsf_params = get_lsf_params(bcftools_stats_summary_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
task = {
'func': bcftools_stats_summary,
'params': {
'in_dir': input_dir,
'out_dir': output_dir,
},
'stage_name': stage,
'uid': 'all-chroms',
'drm_params': lsf_params_json,
'parents': parent_tasks,
}
summary_task = self.workflow.add_task(**task)
return summary_task
def create_concatenate_vcfs_task(self, parent_tasks, step_number):
tasks = list()
stage = self._construct_task_name('concat-vcfs', step_number)
output_dir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(concatenate_vcfs_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
def region_key(task):
reference_fai = os.path.join(
'/gscmnt/ams1102/info/model_data/2869585698/build106942997',
'all_sequences.fa.fai')
return Region(all_sequences.fa.fai, task.params['in_chrom'])
def chromosome_key(task):
reference_fai = os.path.join(
'/gscmnt/ams1102/info/model_data/2869585698/build106942997',
'all_sequences.fa.fai')
return Region(reference_fai, task.params['in_chrom']).chrom
for ref_chrom, chrom_tasks in groupby(sorted(parent_tasks,
key=region_key),
key=chromosome_key):
ptasks = list(chrom_tasks)
input_vcfs = [x.params['out_vcf'] for x in ptasks]
output_vcf = 'concatenated.c{}.vcf.gz'.format(ref_chrom)
output_log = 'concatenate.{}.log'.format(ref_chrom)
task = {
'func': concatenate_vcfs,
'params': {
'in_vcfs': input_vcfs,
'in_chrom': ref_chrom,
'out_vcf': os.path.join(output_dir, ref_chrom, output_vcf),
'out_log': os.path.join(output_dir, ref_chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=ref_chrom),
'drm_params': lsf_params_json,
'parents': ptasks,
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_variant_eval_summary_task(self, parent_tasks, step_number):
stage = self._construct_task_name('gatk-variant-eval-summary',
step_number)
output_dir = os.path.join(self.config.rootdir, stage)
prior_stage_name = parent_tasks[0].stage.name
input_dir = os.path.join(self.config.rootdir, prior_stage_name)
lsf_params = get_lsf_params(variant_eval_summary_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
task = {
'func': variant_eval_summary,
'params': {
'in_dir': input_dir,
'out_dir': output_dir,
},
'stage_name': stage,
'uid': 'all-chroms',
'drm_params': lsf_params_json,
'parents': parent_tasks,
}
summary_task = self.workflow.add_task(**task)
return summary_task
def create_bcftools_stats_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('bcftools-stats', step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(bcftools_stats_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_stats = '{}.stats.out'.format(chrom)
task = {
'func': bcftools_stats,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_stats': os.path.join(basedir, chrom, output_stats),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_variant_eval_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('gatk-variant-eval', step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(gatk_variant_eval_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_stats = 'chrom-{}-variant-eval.out'.format(chrom)
output_log = 'chrom-{}-variant-eval.log'.format(chrom)
task = {
'func': gatk_variant_eval,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_stats': os.path.join(basedir, chrom, output_stats),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_vep_cadd_annotation_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('vep-cadd-annotation', step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(annotation_vep_cadd_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'annotated.vep.cadd.c{}.vcf.gz'.format(chrom)
output_log = 'vep.cadd.annotation.{}.log'.format(chrom)
task = {
'func': annotation_vep_cadd,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_ExAC_annotation_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('annotate-w-ExAC', step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(annotation_ExAC_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'ExAC-annotated.c{}.vcf.gz'.format(chrom)
output_log = 'ExAC-annotate.{}.log'.format(chrom)
task = {
'func': annotation_ExAC,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_1000G_annotation_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('annotate-w-1000G', step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(annotation_1000G_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = '1kg-annotated.c{}.vcf.gz'.format(chrom)
output_log = '1000G-annotate.{}.log'.format(chrom)
task = {
'func': annotation_1000G,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_allele_balance_annotation_tasks(self, parent_tasks,
step_number):
tasks = []
stage = self._construct_task_name('allele-balance-annotation',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(annotate_allele_balances_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'combined.c{chrom}.vcf.gz'.format(chrom=chrom)
output_log = 'allele-balance-{}.log'.format(chrom)
task = {
'func': annotate_allele_balances,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_filter_variant_missingness_tasks(self, parent_tasks,
step_number):
tasks = []
stage = self._construct_task_name('filter-variant-missingness',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(filter_variant_missingness_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'combined.c{chrom}.vcf.gz'.format(chrom=chrom)
output_log = 'filter-missingness-{}.log'.format(chrom)
task = {
'func': filter_variant_missingness,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_remove_symbolic_deletion_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('remove-symbolic-alleles',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(
remove_symbolic_deletion_alleles_lsf_params, self.config.email,
self.config.docker)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'combined.c{chrom}.vcf.gz'.format(chrom=chrom)
output_log = 'remove-symbolic-alleles-chrom-{}.log'.format(chrom)
task = {
'func': remove_symbolic_deletion_alleles,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_decompose_normalize_unique_tasks(self, parent_tasks,
step_number):
tasks = []
stage = self._construct_task_name('decompose-normalize-uniq',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(normalize_decompose_unique_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
output_vcf = 'combined.c{chrom}.vcf.gz'.format(chrom=chrom)
output_log = 'decompose-normalize-unique-{}.log'.format(chrom)
task = {
'func': normalize_decompose_unique,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_calculate_sample_missingness_task(self, parent_tasks,
step_number):
stage = self._construct_task_name('calculate-sample-missingness',
step_number)
output_dir = os.path.join(self.config.rootdir, stage)
prior_stage_name = parent_tasks[0].stage.name
input_dir = os.path.join(self.config.rootdir, prior_stage_name)
input_json_wildcard_path = os.path.join(input_dir, '*', '*.json')
lsf_params = get_lsf_params(calculate_sample_missingness_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
task = {
'func': calculate_sample_missingness,
'params': {
'in_json':
input_json_wildcard_path,
'out_stats':
os.path.join(output_dir, 'sample-missingness-pct.dat'),
'out_log':
os.path.join(output_dir, 'sample-missingness-pct.dat.log'),
},
'stage_name': stage,
'uid': '1-22',
'drm_params': lsf_params_json,
'parents': parent_tasks,
}
summary_task = self.workflow.add_task(**task)
return summary_task
def create_count_sample_missingness_tasks(self, parent_tasks, step_number):
tasks = []
stage = self._construct_task_name('count-sample-missingness',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(count_sample_missingness_lsf_params,
self.config.email, self.config.docker,
self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for ptask in parent_tasks:
chrom = ptask.params['in_chrom']
# only count missing genotypes on chromosomes 1-22 (not X, Y, or MT)
if not chrom[0].isdigit(): continue
output_json = '{chrom}-sample-missingness-counts.json'.format(
chrom=chrom)
output_log = '{}-sample-missingness-counts.log'.format(chrom)
task = {
'func': count_sample_missingness,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_chrom': chrom,
'out_json': os.path.join(basedir, chrom, output_json),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_remove_ac_0_tasks(self, step_number):
tasks = []
stage = self._construct_task_name('select-variants-ac-0-removal',
step_number)
basedir = os.path.join(self.config.rootdir, stage)
lsf_params = get_lsf_params(
gatk_select_variants_remove_ac_0_lsf_params, self.config.email,
self.config.docker, self.config.drm_queue)
lsf_params_json = to_json(lsf_params)
for chrom in self.config.chroms:
vcf = self.config.vcfs[chrom]
output_vcf = 'combined.c{chrom}.vcf.gz'.format(chrom=chrom)
output_log = 'select-variants-chrom-{}-gatk.log'.format(chrom)
task = {
'func': gatk_select_variants_remove_ac_0,
'params': {
'in_chrom': chrom,
'in_vcf': vcf,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'out_log': os.path.join(basedir, chrom, output_log),
},
'stage_name': stage,
'uid': '{chrom}'.format(chrom=chrom),
'drm_params': lsf_params_json,
}
tasks.append(self.workflow.add_task(**task))
return tasks
def _construct_task_name(self, name, number):
task_name = '{}-{}'.format(number, name)
return task_name
def main():
args = parse_args()
cosmos = Cosmos(
"sqlite:///%s/sqlite.db" % os.path.dirname(os.path.abspath(__file__)),
default_drm="awsbatch",
default_drm_options=dict(
container_image=args.container_image,
s3_prefix_for_command_script_temp_files=args.s3_prefix_for_command_script_temp_files,
# only retry on spot instance death
retry_only_if_status_reason_matches="Host EC2 .+ terminated.",
),
default_queue=args.default_queue,
)
cosmos.initdb()
# sp.check_call("mkdir -p analysis_output/ex1", shell=True)
# os.chdir("analysis_output/ex1")
workflow = cosmos.start(f"Evaluate_{args.id}", restart=True, skip_confirm=True)
parameters = np.load(f"optimize_awsbatch/parameters/{args.id}.npy")
for i, par in enumerate(parameters):
parameters_ = dict(
mean_weight=par[0],
c_w=par[1],
tau_pos=par[2],
tau_neg=par[3],
A_pos=par[4],
A_neg=par[5],
weight_decay=par[6],
n_filters=25,
time_max=250,
crop=20,
kernel_size=16,
stride=4,
intensity=127.5,
c_w_min=None,
c_l=True,
network_type="LC_SNN",
)
workflow.add_task(
func=evaluate,
params=dict(
parameters=parameters_,
out_s3_uri=f"{args.out_s3_uri}/scores/{args.id}/{i}.json",
sleep=args.sleep,
train=args.train,
calibrate=args.calibrate,
test=args.test
),
uid=str(i),
time_req=None,
max_attempts=args.max_attempts,
core_req=args.core_req,
mem_req=args.mem_req,
)
workflow.run()
sys.exit(0 if workflow.successful else 1)
class Pipeline(object):
def __init__(self, config, drm, restart):
self.config = config
self.cosmos = Cosmos(database_url='sqlite:///{}'.format(
self.config.db),
get_submit_args=default_get_submit_args,
default_drm=drm)
self.cosmos.initdb()
primary_logfile = os.path.join(
self.config.rootdir,
'{}.log'.format(self.config.project_name),
)
self.workflow = self.cosmos.start(
self.config.project_name,
primary_log_path=primary_logfile,
restart=restart,
)
self.setup_pipeline()
def setup_pipeline(self):
self.construct_pipeline()
self.workflow.make_output_dirs()
def run(self):
# put set_successful to False if you intend to add more tasks to the
# pipeline later
custom_log_dir = lambda task: os.path.join(self.config.rootdir, 'logs',
task.stage.name, task.uid)
self.workflow.run(set_successful=False,
log_out_dir_func=custom_log_dir)
def construct_pipeline(self):
partition_tasks = self.create_vcf_partition_tasks()
plink_pipeline_tasks = self.create_plink_pipeline_tasks(
partition_tasks)
aggregate_mie_stats_tasks = self.create_aggregate_mie_stats_tasks(
plink_pipeline_tasks)
def create_aggregate_mie_stats_tasks(self, parent_tasks):
tasks = []
stage = '3-aggregate-mie-stats'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
input_dir = os.path.join(self.config.rootdir,
parent_tasks[0].stage.name)
parent_snp_tranche_tasks = \
[task for task in parent_tasks \
if (task.params['type'] == 'snps' and task.params['method'] == 'tranche') ]
parent_snp_percentile_tasks = \
[task for task in parent_tasks \
if (task.params['type'] == 'snps' and task.params['method'] == 'percentile') ]
parent_indel_tranche_tasks = \
[task for task in parent_tasks \
if (task.params['type'] == 'indels' and task.params['method'] == 'tranche') ]
parent_indel_percentile_tasks = \
[task for task in parent_tasks \
if (task.params['type'] == 'indels' and task.params['method'] == 'percentile') ]
task_groups = (parent_snp_tranche_tasks, parent_snp_percentile_tasks,
parent_indel_tranche_tasks,
parent_indel_percentile_tasks)
for tgroup in task_groups:
category = tgroup[0].params['type']
method = tgroup[0].params['method']
out_filename = '.'.join([category, method, 'tsv'])
output_file = os.path.join(basedir, out_filename)
task = {
'func':
aggregate_mie_statistics,
'params': {
'in_category': category,
'in_method': method,
'in_dir': input_dir,
'out_file': output_file,
},
'stage_name':
stage,
'uid':
'{category}:{method}'.format(method=method, category=category),
'drm_params':
to_json(aggregate_mie_statistics_lsf_params(email)),
'parents':
tgroup,
}
tasks.append(self.workflow.add_task(**task))
def create_plink_pipeline_tasks(self, parent_tasks):
tasks = []
stage = '2-plink-pipeline'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
for ptask in sorted(parent_tasks, key=lambda t: t.id):
chrom = ptask.params['in_chrom']
label = ptask.params['in_label']
method = ptask.params['in_method']
category = ptask.params['in_type']
output_dir = os.path.join(basedir, category, method, label, chrom)
# ensure_directory(output_dir)
task = {
'func':
plink_pipeline,
'params': {
'in_vcf': ptask.params['out_vcf'],
'in_trio_fam': self.config.plink_fam_file,
'chrom': chrom,
'type': category,
'method': method,
'chrom': chrom,
'label': label,
'out_dir': output_dir,
},
'stage_name':
stage,
'uid':
'{category}:{method}:{label}:{chrom}'.format(chrom=chrom,
method=method,
category=category,
label=label),
'drm_params':
to_json(plink_pipeline_lsf_params(email)),
'parents': [ptask],
}
tasks.append(self.workflow.add_task(**task))
return tasks
def create_vcf_partition_tasks(self):
all_tasks = []
cases = (
('tranche', 'snps', self.config.tranche_intervals['snps']),
('tranche', 'indels', self.config.tranche_intervals['indels']),
('percentile', 'snps', self.config.percentiles['snps']),
('percentile', 'indels', self.config.percentiles['indels']),
)
for case in cases:
tasks = self.generate_vcf_partition_tasks(*case)
all_tasks.extend(tasks)
return all_tasks
def generate_vcf_partition_tasks(self, method, category, intervals):
# method: 'tranche' or 'percentile'
# category: 'snps' or 'indels'
# label:
# tranche : 1, 2 or 3
# percentile : 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
tasks = []
for label in sorted(intervals.keys()):
interval = intervals[label]
partition_tasks = self.create_vcf_partition_chromosome_tasks(
method=method,
label=str(label),
category=category,
interval=interval,
)
tasks.extend(partition_tasks)
return tasks
def create_vcf_partition_chromosome_tasks(self, method, label, category,
interval):
tasks = []
stage = '1-partition-vcfs'
basedir = os.path.join(self.config.rootdir, stage, category, method,
label)
email = self.config.email
for chrom in self.config.chroms:
vcf = self.config.vcfs[chrom]
output_vcf = 'selected.c{chrom}.vcf.gz'.format(chrom=chrom)
task = {
'func':
vcf_partition,
'params': {
'in_vcf': vcf,
'out_vcf': os.path.join(basedir, chrom, output_vcf),
'in_min_vqslod': interval[0],
'in_max_vqslod': interval[1],
'in_samples': self.config.control_samples_file,
'in_type': category,
'in_method': method,
'in_chrom': chrom,
'in_label': label,
},
'stage_name':
stage,
'uid':
'{category}:{method}:{label}:{chrom}'.format(chrom=chrom,
method=method,
category=category,
label=label),
'drm_params':
to_json(vcf_partition_lsf_params(email)),
}
tasks.append(self.workflow.add_task(**task))
return tasks
class Pipeline(object):
def __init__(self, config, drm, restart):
self.config = config
self.cosmos = Cosmos(
database_url='sqlite:///{}'.format(self.config.db),
get_submit_args=default_get_submit_args,
default_drm=drm
)
self.cosmos.initdb()
primary_logfile = os.path.join(
self.config.rootdir,
'{}.log'.format(self.config.project_name),
)
self.workflow = self.cosmos.start(
self.config.project_name,
primary_log_path=primary_logfile,
restart=restart,
)
self.setup_pipeline()
def setup_pipeline(self):
self.construct_pipeline()
self.workflow.make_output_dirs()
def run(self):
# put set_successful to False if you intend to add more tasks to the
# pipeline later
custom_log_dir = lambda task : os.path.join(self.config.rootdir, 'logs', task.stage.name, task.uid)
self.workflow.run(set_successful=False, log_out_dir_func=custom_log_dir)
def construct_pipeline(self):
filter_biallelic_snps_tasks = self.create_filter_biallelic_snps_tasks()
plink_binary_tasks = self.create_plink_binary_tasks(filter_biallelic_snps_tasks)
plink_ld_prune_tasks = self.create_plink_ld_prune_tasks(plink_binary_tasks)
plink_extract_prune_tasks = self.create_plink_extract_prune_tasks(plink_ld_prune_tasks)
plink_merge_prune_files_task = self.create_plink_merge_prune_file_task(plink_extract_prune_tasks)
eigenstrat_task = self.create_eigenstrat_smartpca_task(plink_merge_prune_files_task)
data_frame_task = self.create_data_frame_task(eigenstrat_task)
def create_data_frame_task(self, parent_task):
stage = '7-make-data-frame'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
pca_evec_file = os.path.join(
parent_task.params['out_prj_dir'],
'merged.eigenstrat.pca.evec',
)
out_file = os.path.join(basedir, 'merged.eigenstrat.pca.evec.tsv')
task = {
'func' : create_evec_data_frame,
'params' : {
'in_file' : pca_evec_file,
'out_file' : out_file,
},
'stage_name' : stage,
'uid' : 'all-chroms',
'drm_params' :
to_json(create_evec_data_frame_lsf_params(email)),
'parents' : [ parent_task ],
}
df_task = self.workflow.add_task(**task)
return df_task
def create_eigenstrat_smartpca_task(self, parent_task):
stage = '6-eigenstrat-smartpca'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
ped_file = "{}.ped".format(parent_task.params['out_path'])
map_file = "{}.map".format(parent_task.params['out_path'])
task = {
'func' : eigenstrat_smartpca_analysis,
'params' : {
'in_ped_file' : ped_file,
'in_map_file' : map_file,
'out_prj_dir' : basedir,
},
'stage_name' : stage,
'uid' : 'all-chroms',
'drm_params' :
to_json(eigenstrat_smartpca_analysis_lsf_params(email)),
'parents' : [ parent_task ],
}
eigenstrat_task = self.workflow.add_task(**task)
return eigenstrat_task
def create_plink_merge_prune_file_task(self, parent_tasks):
stage = '5-plink-merge-prune-files'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
parent_tasks_sorted = sorted(parent_tasks, key=lambda t: t.id)
first_task = parent_tasks_sorted[0]
remaining_tasks = parent_tasks_sorted[1:]
merge_list_file = os.path.join(basedir, 'allfiles.txt')
self._create_merge_list(merge_list_file, remaining_tasks)
output_path = os.path.join(basedir, 'merged')
task = {
'func' : plink_merge_pruned_files,
'params' : {
'in_ref' : first_task.params['out_path'],
'in_merge_file' : merge_list_file,
'out_path' : output_path,
},
'stage_name' : stage,
'uid' : 'all-chroms',
'drm_params' :
to_json(plink_merge_pruned_files_lsf_params(email)),
'parents' : parent_tasks_sorted,
}
merge_task = self.workflow.add_task(**task)
return merge_task
def _create_merge_list(self, merge_file, tasks):
ensure_directory(os.path.dirname(merge_file))
with open(merge_file, 'w') as f:
for t in tasks:
print(t.params['out_path'], file=f)
def create_plink_extract_prune_tasks(self, parent_tasks):
tasks = []
stage = '4-plink-extract-prune'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
for ptask in sorted(parent_tasks, key=lambda t: t.id):
plink_extract_file = "{}.prune.in".format(ptask.params['out_path'])
orig_binary_data = ptask.params['in_path']
chrom = ptask.params['chrom']
output_path = os.path.join(basedir, chrom, 'c{}.extracted'.format(chrom))
task = {
'func' : plink_extract_prune,
'params' : {
'in_path' : orig_binary_data,
'in_extract' : plink_extract_file,
'out_path' : output_path,
'chrom' : chrom,
},
'stage_name' : stage,
'uid' : '{chrom}'.format(chrom=chrom),
'drm_params' :
to_json(plink_extract_prune_lsf_params(email)),
'parents' : [ptask],
}
tasks.append( self.workflow.add_task(**task) )
return tasks
def create_plink_ld_prune_tasks(self, parent_tasks):
tasks = []
stage = '3-plink-ld-prune'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
for ptask in sorted(parent_tasks, key=lambda t: t.id):
chrom = ptask.params['chrom']
output_path = os.path.join(basedir, chrom, 'c{}-pruned'.format(chrom))
task = {
'func' : plink_ld_prune,
'params' : {
'in_path' : ptask.params['out_path'],
'out_path' : output_path,
'chrom' : chrom,
},
'stage_name' : stage,
'uid' : '{chrom}'.format(chrom=chrom),
'drm_params' :
to_json(plink_ld_prune_lsf_params(email)),
'parents' : [ptask],
}
tasks.append( self.workflow.add_task(**task) )
return tasks
def create_plink_binary_tasks(self, parent_tasks):
tasks = []
stage = '2-plink-binaries'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
for ptask in sorted(parent_tasks, key=lambda t: t.id):
chrom = ptask.params['chrom']
output_path = os.path.join(basedir, chrom, 'c{}'.format(chrom))
task = {
'func' : plink_binary,
'params' : {
'in_vcf' : ptask.params['out_vcf'],
'out_path' : output_path,
'chrom' : chrom,
},
'stage_name' : stage,
'uid' : '{chrom}'.format(chrom=chrom),
'drm_params' :
to_json(plink_binary_lsf_params(email)),
'parents' : [ptask],
}
tasks.append( self.workflow.add_task(**task) )
return tasks
def create_filter_biallelic_snps_tasks(self):
tasks = []
stage = '1-filter-biallelic-snps'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
for chrom in self.config.chroms:
vcf = self.config.vcfs[chrom]
output_vcf = 'filtered.snps.c{chrom}.vcf.gz'.format(chrom=chrom)
task = {
'func' : filter_biallelic_snps,
'params' : {
'chrom' : chrom,
'in_vcf' : vcf,
'out_vcf' : os.path.join(basedir, chrom, output_vcf),
'in_min_vqslod' : self.config.vqslod_threshold,
},
'stage_name' : stage,
'uid' : '{chrom}'.format(chrom=chrom),
'drm_params' :
to_json(filter_biallelic_snps_lsf_params(email)),
}
tasks.append( self.workflow.add_task(**task) )
return tasks
class Pipeline(object):
def __init__(self, config, drm, restart):
self.config = config
self.cosmos = Cosmos(
database_url='sqlite:///{}'.format(self.config.db),
get_submit_args=default_get_submit_args,
default_drm=drm
)
self.cosmos.initdb()
primary_logfile = os.path.join(
self.config.rootdir,
'{}.log'.format(self.config.project_name),
)
self.workflow = self.cosmos.start(
self.config.project_name,
primary_log_path=primary_logfile,
restart=restart,
)
self.setup_pipeline()
def setup_pipeline(self):
self.construct_pipeline()
self.workflow.make_output_dirs()
def run(self):
# put set_successful to False if you intend to add more tasks to the
# pipeline later
custom_log_dir = lambda task : os.path.join(self.config.rootdir, 'logs', task.stage.name, task.uid)
self.workflow.run(set_successful=False, log_out_dir_func=custom_log_dir)
def construct_pipeline(self):
speedseq_tasks = self.create_speedseq_realign_tasks()
def create_speedseq_realign_tasks(self):
tasks = []
stage = '1-exec-speedseq-realign'
basedir = os.path.join(self.config.rootdir, stage)
email = self.config.email
lsf_job_group = self.config.drm_job_group
sample_data = self.config.sample_data
for sample_id in sample_data.keys():
bam_paths = sample_data[sample_id]['bams']
sample_name = sample_data[sample_id]['meta']['original-name']
output_prefix = os.path.join(basedir, sample_id, "{}.b38.realign".format(sample_id))
tmpdir = os.path.join(basedir, sample_id, 'tmpdir')
input_bams = ' '.join(bam_paths)
task = {
'func' : exec_speedseq,
'params' : {
'output_prefix' : output_prefix,
'tmpdir' : tmpdir,
'input_bams' : input_bams,
},
'stage_name' : stage,
'uid' : sample_id,
'drm_params' :
to_json(exec_speedseq_lsf_params(email, lsf_job_group)),
}
tasks.append( self.workflow.add_task(**task) )
return tasks
