def main():
args = parse_args()
# load and merge peaks
merged_peaks, signal_type = load_samples(args)
s1 = numpy.array([pk.signals[0] for pk in merged_peaks])
s2 = numpy.array([pk.signals[1] for pk in merged_peaks])
# build the ranks vector
idr.log("Ranking peaks", 'VERBOSE')
r1, r2 = build_rank_vectors(merged_peaks)
if args.only_merge_peaks:
localIDRs, IDRs = None, None
else:
if len(merged_peaks) < 20:
error_msg = "Peak files must contain at least 20 peaks post-merge"
error_msg += "\nHint: Merged peaks were written to the output file"
write_results_to_file(
merged_peaks, args.output_file,
args.output_file_type, signal_type)
raise ValueError(error_msg)
localIDRs = fit_model_and_calc_local_idr(
r1, r2,
starting_point=(
args.initial_mu, args.initial_sigma,
args.initial_rho, args.initial_mix_param),
max_iter=args.max_iter,
convergence_eps=args.convergence_eps,
fix_mu=args.fix_mu, fix_sigma=args.fix_sigma )
if args.use_best_multisummit_IDR:
localIDRs = correct_multi_summit_peak_IDR_values(
localIDRs, merged_peaks)
IDRs = calc_global_IDR(localIDRs)
if args.plot:
assert len(args.samples) == 2
plot(args, [s1, s2], [r1, r2], IDRs)
num_peaks_passing_thresh = write_results_to_file(
merged_peaks,
args.output_file,
args.output_file_type,
signal_type,
localIDRs=localIDRs,
IDRs=IDRs,
max_allowed_idr=args.idr_threshold,
soft_max_allowed_idr=args.soft_idr_threshold,
useBackwardsCompatibleOutput=args.use_old_output_format)
args.output_file.close()
def main():
args = parse_args()
# load and merge peaks
merged_peaks, signal_type = load_samples(args)
s1 = numpy.array([pk.signals[0] for pk in merged_peaks])
s2 = numpy.array([pk.signals[1] for pk in merged_peaks])
# build the ranks vector
idr.log("Ranking peaks", 'VERBOSE')
r1, r2 = build_rank_vectors(merged_peaks)
if args.only_merge_peaks:
localIDRs, IDRs = None, None
else:
if len(merged_peaks) < 20:
error_msg = "Peak files must contain at least 20 peaks post-merge"
error_msg += "\nHint: Merged peaks were written to the output file"
write_results_to_file(merged_peaks, args.output_file,
args.input_file_type, signal_type)
raise ValueError(error_msg)
localIDRs, IDRs = fit_model_and_calc_idr(
r1,
r2,
starting_point=(args.initial_mu, args.initial_sigma,
args.initial_rho, args.initial_mix_param),
max_iter=args.max_iter,
convergence_eps=args.convergence_eps,
fix_mu=args.fix_mu,
fix_sigma=args.fix_sigma)
if args.plot:
assert len(args.samples) == 2
plot(args, [s1, s2], [r1, r2], IDRs)
num_peaks_passing_thresh = write_results_to_file(
merged_peaks,
args.output_file,
args.input_file_type,
signal_type,
localIDRs=localIDRs,
IDRs=IDRs,
max_allowed_idr=args.idr_threshold,
soft_max_allowed_idr=args.soft_idr_threshold,
useBackwardsCompatibleOutput=args.use_old_output_format)
args.output_file.close()
def fit_model_and_calc_idr(r1,
r2,
starting_point=None,
max_iter=idr.MAX_ITER_DEFAULT,
convergence_eps=idr.CONVERGENCE_EPS_DEFAULT,
fix_mu=False,
fix_sigma=False):
# in theory we would try to find good starting point here,
# but for now just set it to somethign reasonable
if type(starting_point) == type(None):
starting_point = (DEFAULT_MU, DEFAULT_SIGMA, DEFAULT_RHO,
DEFAULT_MIX_PARAM)
idr.log("Initial parameter values: [%s]" %
" ".join("%.2f" % x for x in starting_point))
# fit the model parameters
idr.log("Fitting the model parameters", 'VERBOSE')
if idr.PROFILE:
import cProfile
cProfile.runctx(
"""theta, loss = estimate_model_params(
r1,r2,
starting_point,
max_iter=max_iter,
convergence_eps=convergence_eps,
fix_mu=fix_mu, fix_sigma=fix_sigma)
""",
{'estimate_model_params': estimate_model_params}, {
'r1': r1,
'r2': r2,
'starting_point': starting_point,
'max_iter': max_iter,
'convergence_eps': convergence_eps,
'fix_mu': fix_mu,
'fix_sigma': fix_sigma
})
assert False
theta, loss = estimate_model_params(r1,
r2,
starting_point,
max_iter=max_iter,
convergence_eps=convergence_eps,
fix_mu=fix_mu,
fix_sigma=fix_sigma)
idr.log("Finished running IDR on the datasets", 'VERBOSE')
idr.log("Final parameter values: [%s]" % " ".join("%.2f" % x
for x in theta))
# calculate the global IDR
localIDRs, IDRs = calc_IDR(numpy.array(theta), r1, r2)
return localIDRs, IDRs
def fit_model_and_calc_idr(r1, r2,
starting_point=None,
max_iter=idr.MAX_ITER_DEFAULT,
convergence_eps=idr.CONVERGENCE_EPS_DEFAULT,
fix_mu=False, fix_sigma=False ):
# in theory we would try to find good starting point here,
# but for now just set it to somethign reasonable
if type(starting_point) == type(None):
starting_point = (DEFAULT_MU, DEFAULT_SIGMA,
DEFAULT_RHO, DEFAULT_MIX_PARAM)
idr.log("Initial parameter values: [%s]" % " ".join(
"%.2f" % x for x in starting_point))
# fit the model parameters
idr.log("Fitting the model parameters", 'VERBOSE');
if idr.PROFILE:
import cProfile
cProfile.runctx("""theta, loss = estimate_model_params(
r1,r2,
starting_point,
max_iter=max_iter,
convergence_eps=convergence_eps,
fix_mu=fix_mu, fix_sigma=fix_sigma)
""",
{'estimate_model_params': estimate_model_params},
{'r1':r1, 'r2':r2,
'starting_point': starting_point,
'max_iter': max_iter,
'convergence_eps': convergence_eps,
'fix_mu': fix_mu, 'fix_sigma': fix_sigma} )
assert False
theta, loss = estimate_model_params(
r1, r2,
starting_point,
max_iter=max_iter,
convergence_eps=convergence_eps,
fix_mu=fix_mu, fix_sigma=fix_sigma)
idr.log("Finished running IDR on the datasets", 'VERBOSE')
idr.log("Final parameter values: [%s]"%" ".join("%.2f" % x for x in theta))
# calculate the global IDR
localIDRs, IDRs = calc_IDR(numpy.array(theta), r1, r2)
return localIDRs, IDRs
def write_results_to_file(merged_peaks, output_file,
output_file_type, signal_type,
max_allowed_idr=1.0,
soft_max_allowed_idr=1.0,
localIDRs=None, IDRs=None,
useBackwardsCompatibleOutput=False):
if useBackwardsCompatibleOutput:
build_idr_output_line = build_backwards_compatible_idr_output_line
else:
build_idr_output_line = build_idr_output_line_with_bed6
# write out the result
idr.log("Writing results to file", "VERBOSE");
if localIDRs is None or IDRs is None:
assert IDRs is None
assert localIDRs is None
localIDRs = numpy.ones(len(merged_peaks))
IDRs = numpy.ones(len(merged_peaks))
num_peaks_passing_hard_thresh = 0
num_peaks_passing_soft_thresh = 0
for localIDR, IDR, merged_peak in zip(
localIDRs, IDRs, merged_peaks):
# skip peaks with global idr values below the threshold
if max_allowed_idr != None and IDR > max_allowed_idr:
continue
num_peaks_passing_hard_thresh += 1
if IDR <= soft_max_allowed_idr:
num_peaks_passing_soft_thresh += 1
opline = build_idr_output_line(
merged_peak, IDR, localIDR, output_file_type, signal_type)
print( opline, file=output_file )
if len(merged_peaks) == 0: return
idr.log(
"Number of reported peaks - {}/{} ({:.1f}%)\n".format(
num_peaks_passing_hard_thresh, len(merged_peaks),
100*float(num_peaks_passing_hard_thresh)/len(merged_peaks))
)
idr.log(
"Number of peaks passing IDR cutoff of {} - {}/{} ({:.1f}%)\n".format(
soft_max_allowed_idr,
num_peaks_passing_soft_thresh, len(merged_peaks),
100*float(num_peaks_passing_soft_thresh)/len(merged_peaks))
)
return
def write_results_to_file(merged_peaks, output_file,
output_file_type, signal_type,
max_allowed_idr=1.0,
soft_max_allowed_idr=1.0,
localIDRs=None, IDRs=None,
useBackwardsCompatibleOutput=False):
if useBackwardsCompatibleOutput:
build_idr_output_line = build_backwards_compatible_idr_output_line
else:
build_idr_output_line = build_idr_output_line_with_bed6
# write out the result
idr.log("Writing results to file", "VERBOSE");
if localIDRs is None or IDRs is None:
assert IDRs is None
assert localIDRs is None
localIDRs = numpy.ones(len(merged_peaks))
IDRs = numpy.ones(len(merged_peaks))
num_peaks_passing_hard_thresh = 0
num_peaks_passing_soft_thresh = 0
for localIDR, IDR, merged_peak in zip(
localIDRs, IDRs, merged_peaks):
# skip peaks with global idr values below the threshold
if max_allowed_idr != None and IDR > max_allowed_idr:
continue
num_peaks_passing_hard_thresh += 1
if IDR <= soft_max_allowed_idr:
num_peaks_passing_soft_thresh += 1
opline = build_idr_output_line(
merged_peak, IDR, localIDR, output_file_type, signal_type)
print( opline, file=output_file )
if len(merged_peaks) == 0: return
idr.log(
"Number of reported peaks - {}/{} ({:.1f}%)\n".format(
num_peaks_passing_hard_thresh, len(merged_peaks),
100*float(num_peaks_passing_hard_thresh)/len(merged_peaks))
)
idr.log(
"Number of peaks passing IDR cutoff of {} - {}/{} ({:.1f}%)\n".format(
soft_max_allowed_idr,
num_peaks_passing_soft_thresh, len(merged_peaks),
100*float(num_peaks_passing_soft_thresh)/len(merged_peaks))
)
return
def load_samples(args):
# decide what aggregation function to use for peaks that need to be merged
idr.log("Loading the peak files", 'VERBOSE')
if args.input_file_type in ['narrowPeak', 'broadPeak']:
if args.rank == None: signal_type = 'signal.value'
else: signal_type = args.rank
try:
signal_index = {
"score": 4,
"signal.value": 6,
"p.value": 7,
"q.value": 8
}[signal_type]
except KeyError:
raise ValueError(
"Unrecognized signal type for {} filetype: '{}'".format(
args.input_file_type, signal_type))
if args.peak_merge_method != None:
peak_merge_fn = {
"sum": sum,
"avg": mean,
"min": min,
"max": max
}[args.peak_merge_method]
elif signal_index in (4, 6):
peak_merge_fn = sum
else:
peak_merge_fn = min
if args.input_file_type == 'narrowPeak':
summit_index = 9
else:
summit_index = None
f1, f2 = [
load_bed(fp, signal_index, summit_index) for fp in args.samples
]
oracle_pks = (load_bed(args.peak_list, signal_index, summit_index)
if args.peak_list != None else None)
elif args.input_file_type in [
'bed',
]:
# set the default
if args.rank == None:
signal_type = 'score'
if args.rank == 'score':
signal_type = 'score'
signal_index = 4
else:
try:
signal_index = int(args.rank) - 1
signal_type = "COL%i" % (signal_index + 1)
except ValueError:
raise ValueError("For bed files --signal-type must either "\
+"be set to score or an index specifying "\
+"the column to use.")
if args.peak_merge_method == None:
peak_merge_fn = sum
else:
peak_merge_fn = {
"sum": sum,
"avg": mean,
"min": min,
"max": max
}[args.peak_merge_method]
f1, f2 = [load_bed(fp, signal_index) for fp in args.samples]
oracle_pks = (load_bed(args.peak_list, signal_index)
if args.peak_list != None else None)
else:
raise ValueError("Unrecognized file type: '{}'".format(
args.input_file_type))
# build a unified peak set
idr.log("Merging peaks", 'VERBOSE')
merged_peaks = merge_peaks([f1, f2], peak_merge_fn, oracle_pks,
args.use_nonoverlapping_peaks)
return merged_peaks, signal_type
def parse_args():
import argparse
parser = argparse.ArgumentParser(
formatter_class=argparse.RawTextHelpFormatter,
description="""
Program: IDR (Irreproducible Discovery Rate)
Version: {PACKAGE_VERSION}
Contact: Nathan Boley <*****@*****.**>
""".format(PACKAGE_VERSION=idr.__version__))
def PossiblyGzippedFile(fname):
if fname.endswith(".gz"):
return io.TextIOWrapper(gzip.open(fname, 'rb'))
else:
return open(fname, 'r')
parser.add_argument('--samples',
'-s',
type=PossiblyGzippedFile,
nargs=2,
required=True,
help='Files containing peaks and scores.')
parser.add_argument(
'--peak-list',
'-p',
type=PossiblyGzippedFile,
help='If provided, all peaks will be taken from this file.')
parser.add_argument('--input-file-type',
default='narrowPeak',
choices=['narrowPeak', 'broadPeak', 'bed'],
help='File type of --samples and --peak-list.')
parser.add_argument( '--rank',
help="Which column to use to rank peaks."\
+"\t\nOptions: signal.value p.value q.value columnIndex"\
+"\nDefaults:\n\tnarrowPeak/broadPeak: signal.value\n\tbed: score")
default_ofname = "idrValues.txt"
parser.add_argument(
'--output-file',
"-o",
default=default_ofname,
help='File to write output to.\nDefault: {}'.format(default_ofname))
parser.add_argument('--log-output-file',
"-l",
type=argparse.FileType("w"),
default=sys.stderr,
help='File to write output to. Default: stderr')
parser.add_argument( '--idr-threshold', "-i", type=float, default=None,
help="Only return peaks with a global idr threshold below this value."\
+"\nDefault: report all peaks")
parser.add_argument( '--soft-idr-threshold', type=float, default=None,
help="Report statistics for peaks with a global idr below this "\
+"value but return all peaks with an idr below --idr.\nDefault: %.2f" \
% idr.DEFAULT_SOFT_IDR_THRESH)
parser.add_argument('--use-old-output-format',
action='store_true',
default=False,
help="Use old output format.")
parser.add_argument('--plot',
action='store_true',
default=False,
help='Plot the results to [OFNAME].png')
parser.add_argument(
'--use-nonoverlapping-peaks',
action="store_true",
default=False,
help='Use peaks without an overlapping match and set the value to 0.')
parser.add_argument( '--peak-merge-method',
choices=["sum", "avg", "min", "max"], default=None,
help="Which method to use for merging peaks.\n" \
+ "\tDefault: 'sum' for signal/score/column indexes, 'min' for p/q-value.")
parser.add_argument('--initial-mu',
type=float,
default=idr.DEFAULT_MU,
help="Initial value of mu. Default: %.2f" %
idr.DEFAULT_MU)
parser.add_argument('--initial-sigma',
type=float,
default=idr.DEFAULT_SIGMA,
help="Initial value of sigma. Default: %.2f" %
idr.DEFAULT_SIGMA)
parser.add_argument('--initial-rho',
type=float,
default=idr.DEFAULT_RHO,
help="Initial value of rho. Default: %.2f" %
idr.DEFAULT_RHO)
parser.add_argument( '--initial-mix-param',
type=float, default=idr.DEFAULT_MIX_PARAM,
help="Initial value of the mixture params. Default: %.2f" \
% idr.DEFAULT_MIX_PARAM)
parser.add_argument(
'--fix-mu',
action='store_true',
help="Fix mu to the starting point and do not let it vary.")
parser.add_argument(
'--fix-sigma',
action='store_true',
help="Fix sigma to the starting point and do not let it vary.")
parser.add_argument(
'--random-seed',
type=int,
default=0,
help="The random seed value (sor braking ties). Default: 0")
parser.add_argument(
'--max-iter',
type=int,
default=idr.MAX_ITER_DEFAULT,
help="The maximum number of optimization iterations. Default: %i" %
idr.MAX_ITER_DEFAULT)
parser.add_argument( '--convergence-eps', type=float,
default=idr.CONVERGENCE_EPS_DEFAULT,
help="The maximum change in parameter value changes " \
+ "for convergence. Default: %.2e" % idr.CONVERGENCE_EPS_DEFAULT)
parser.add_argument('--only-merge-peaks',
action='store_true',
help="Only return the merged peak list.")
parser.add_argument('--verbose',
action="store_true",
default=False,
help="Print out additional debug information")
parser.add_argument('--quiet',
action="store_true",
default=False,
help="Don't print any status messages")
parser.add_argument('--version',
action='version',
version='IDR %s' % idr.__version__)
args = parser.parse_args()
args.output_file = open(args.output_file, "w")
idr.log_ofp = args.log_output_file
if args.verbose:
idr.VERBOSE = True
global QUIET
if args.quiet:
idr.QUIET = True
idr.VERBOSE = False
assert idr.DEFAULT_IDR_THRESH == 1.0
if args.idr_threshold == None and args.soft_idr_threshold == None:
args.idr_threshold = idr.DEFAULT_IDR_THRESH
args.soft_idr_threshold = idr.DEFAULT_SOFT_IDR_THRESH
elif args.soft_idr_threshold == None:
assert args.idr_threshold != None
args.soft_idr_threshold = args.idr_threshold
elif args.idr_threshold == None:
assert args.soft_idr_threshold != None
args.idr_threshold = idr.DEFAULT_IDR_THRESH
numpy.random.seed(args.random_seed)
if args.plot:
try:
import matplotlib
except ImportError:
idr.log("WARNING: matplotlib does not appear to be installed and "\
+"is required for plotting - turning plotting off.",
level="WARNING" )
args.plot = False
return args
def load_samples(args):
# decide what aggregation function to use for peaks that need to be merged
idr.log("Loading the peak files", 'VERBOSE')
if args.input_file_type in ['narrowPeak', 'broadPeak']:
if args.rank == None: signal_type = 'signal.value'
else: signal_type = args.rank
try:
signal_index = {"score": 4, "signal.value": 6,
"p.value": 7, "q.value": 8}[signal_type]
except KeyError:
raise ValueError(
"Unrecognized signal type for {} filetype: '{}'".format(
args.input_file_type, signal_type))
if args.peak_merge_method != None:
peak_merge_fn = {
"sum": sum, "avg": mean, "min": min, "max": max}[
args.peak_merge_method]
elif signal_index in (4,6):
peak_merge_fn = sum
else:
peak_merge_fn = min
if args.input_file_type == 'narrowPeak':
summit_index = 9
else:
summit_index = None
f1, f2 = [load_bed(fp, signal_index, summit_index)
for fp in args.samples]
oracle_pks = (
load_bed(args.peak_list, signal_index, summit_index)
if args.peak_list != None else None)
elif args.input_file_type in ['bed', ]:
# set the default
if args.rank == None:
signal_type = 'score'
if args.rank == 'score':
signal_type = 'score'
signal_index = 4
else:
try:
signal_index = int(args.rank) - 1
signal_type = "COL%i" % (signal_index + 1)
except ValueError:
raise ValueError("For bed files --signal-type must either "\
+"be set to score or an index specifying "\
+"the column to use.")
if args.peak_merge_method == None:
peak_merge_fn = sum
else:
peak_merge_fn = {
"sum": sum, "avg": mean, "min": min, "max": max}[
args.peak_merge_method]
f1, f2 = [load_bed(fp, signal_index) for fp in args.samples]
oracle_pks = (
load_bed(args.peak_list, signal_index)
if args.peak_list != None else None)
elif args.input_file_type in ['gff', ]:
# set the default
if args.rank == None:
signal_type = 'score'
else:
assert args.rank == 'score'
if args.peak_merge_method == None:
peak_merge_fn = sum
else:
peak_merge_fn = {
"sum": sum, "avg": mean, "min": min, "max": max}[
args.peak_merge_method]
f1, f2 = [load_gff(fp) for fp in args.samples]
oracle_pks = (
load_gff(args.peak_list)
if args.peak_list != None else None)
else:
raise ValueError( "Unrecognized file type: '{}'".format(
args.input_file_type))
# build a unified peak set
idr.log("Merging peaks", 'VERBOSE')
merged_peaks = merge_peaks([f1, f2], peak_merge_fn,
oracle_pks, args.use_nonoverlapping_peaks)
return merged_peaks, signal_type
def parse_args():
import argparse
parser = argparse.ArgumentParser(
formatter_class=argparse.RawTextHelpFormatter,
description="""
Program: IDR (Irreproducible Discovery Rate)
Version: {PACKAGE_VERSION}
Contact: Nathan Boley <*****@*****.**>
""".format(PACKAGE_VERSION=idr.__version__))
def PossiblyGzippedFile(fname):
if fname.endswith(".gz"):
return io.TextIOWrapper(gzip.open(fname, 'rb'))
else:
return open(fname, 'r')
parser.add_argument( '--samples', '-s', type=PossiblyGzippedFile, nargs=2,
required=True,
help='Files containing peaks and scores.')
parser.add_argument( '--peak-list', '-p', type=PossiblyGzippedFile,
help='If provided, all peaks will be taken from this file.')
parser.add_argument( '--input-file-type', default='narrowPeak',
choices=['narrowPeak', 'broadPeak', 'bed', 'gff'],
help='File type of --samples and --peak-list.')
parser.add_argument( '--rank',
help="Which column to use to rank peaks."\
+"\t\nOptions: signal.value p.value q.value columnIndex"\
+"\nDefaults:\n\tnarrowPeak/broadPeak: signal.value\n\tbed: score")
default_ofname = "idrValues.txt"
parser.add_argument( '--output-file', "-o",
default=default_ofname,
help='File to write output to.\nDefault: {}'.format(default_ofname))
parser.add_argument( '--output-file-type',
choices=['narrowPeak', 'broadPeak', 'bed'],
default=None,
help='Output file type. Defaults to input file type when available, otherwise bed.')
parser.add_argument( '--log-output-file', "-l", type=argparse.FileType("w"),
default=sys.stderr,
help='File to write output to. Default: stderr')
parser.add_argument( '--idr-threshold', "-i", type=float, default=None,
help="Only return peaks with a global idr threshold below this value."\
+"\nDefault: report all peaks")
parser.add_argument( '--soft-idr-threshold', type=float, default=None,
help="Report statistics for peaks with a global idr below this "\
+"value but return all peaks with an idr below --idr.\nDefault: %.2f" \
% idr.DEFAULT_SOFT_IDR_THRESH)
parser.add_argument( '--use-old-output-format',
action='store_true', default=False,
help="Use old output format.")
parser.add_argument( '--plot', action='store_true', default=False,
help='Plot the results to [OFNAME].png')
parser.add_argument( '--use-nonoverlapping-peaks',
action="store_true", default=False,
help='Use peaks without an overlapping match and set the value to 0.')
parser.add_argument( '--peak-merge-method',
choices=["sum", "avg", "min", "max"], default=None,
help="Which method to use for merging peaks.\n" \
+ "\tDefault: 'sum' for signal/score/column indexes, 'min' for p/q-value.")
parser.add_argument( '--initial-mu', type=float, default=idr.DEFAULT_MU,
help="Initial value of mu. Default: %.2f" % idr.DEFAULT_MU)
parser.add_argument( '--initial-sigma', type=float,
default=idr.DEFAULT_SIGMA,
help="Initial value of sigma. Default: %.2f" % idr.DEFAULT_SIGMA)
parser.add_argument( '--initial-rho', type=float, default=idr.DEFAULT_RHO,
help="Initial value of rho. Default: %.2f" % idr.DEFAULT_RHO)
parser.add_argument( '--initial-mix-param',
type=float, default=idr.DEFAULT_MIX_PARAM,
help="Initial value of the mixture params. Default: %.2f" \
% idr.DEFAULT_MIX_PARAM)
parser.add_argument( '--fix-mu', action='store_true',
help="Fix mu to the starting point and do not let it vary.")
parser.add_argument( '--fix-sigma', action='store_true',
help="Fix sigma to the starting point and do not let it vary.")
parser.add_argument( '--dont-filter-peaks-below-noise-mean',
default=False,
action='store_true',
help="Allow signal points that are below the noise mean (should only be used if you know what you are doing).")
parser.add_argument( '--use-best-multisummit-IDR',
default=False, action='store_true',
help="Set the IDR value for a group of multi summit peaks (same chr/start/stop but different summit) to the best value across all peaks. This is a work around for peak callers that don't do a good job splitting scores across multi summit peaks.")
parser.add_argument( '--allow-negative-scores',
default=False,
action='store_true',
help="Allow negative values for scores. (should only be used if you know what you are doing)")
parser.add_argument( '--random-seed', type=int, default=0,
help="The random seed value (sor braking ties). Default: 0")
parser.add_argument( '--max-iter', type=int, default=idr.MAX_ITER_DEFAULT,
help="The maximum number of optimization iterations. Default: %i"
% idr.MAX_ITER_DEFAULT)
parser.add_argument( '--convergence-eps', type=float,
default=idr.CONVERGENCE_EPS_DEFAULT,
help="The maximum change in parameter value changes " \
+ "for convergence. Default: %.2e" % idr.CONVERGENCE_EPS_DEFAULT)
parser.add_argument( '--only-merge-peaks', action='store_true',
help="Only return the merged peak list.")
parser.add_argument( '--verbose', action="store_true", default=False,
help="Print out additional debug information")
parser.add_argument( '--quiet', action="store_true", default=False,
help="Don't print any status messages")
parser.add_argument('--version', action='version',
version='IDR %s' % idr.__version__)
args = parser.parse_args()
args.output_file = open(args.output_file, "w")
idr.log_ofp = args.log_output_file
if args.output_file_type is None:
if args.input_file_type in ('narrowPeak', 'broadPeak', 'bed'):
args.output_file_type = args.input_file_type
else:
args.output_file_type = 'bed'
if args.verbose:
idr.VERBOSE = True
global QUIET
if args.quiet:
idr.QUIET = True
idr.VERBOSE = False
if args.dont_filter_peaks_below_noise_mean is True:
idr.FILTER_PEAKS_BELOW_NOISE_MEAN = False
if args.allow_negative_scores is True:
idr.ONLY_ALLOW_NON_NEGATIVE_VALUES = False
assert idr.DEFAULT_IDR_THRESH == 1.0
if args.idr_threshold == None and args.soft_idr_threshold == None:
args.idr_threshold = idr.DEFAULT_IDR_THRESH
args.soft_idr_threshold = idr.DEFAULT_SOFT_IDR_THRESH
elif args.soft_idr_threshold == None:
assert args.idr_threshold != None
args.soft_idr_threshold = args.idr_threshold
elif args.idr_threshold == None:
assert args.soft_idr_threshold != None
args.idr_threshold = idr.DEFAULT_IDR_THRESH
numpy.random.seed(args.random_seed)
if args.plot:
try:
import matplotlib
except ImportError:
idr.log("WARNING: matplotlib does not appear to be installed and "\
+"is required for plotting - turning plotting off.",
level="WARNING" )
args.plot = False
return args
def main():
args = parse_args()
# load and merge peaks
merged_peaks, signal_type = load_samples(args)
s1 = numpy.array([pk.signals[0] for pk in merged_peaks])
s2 = numpy.array([pk.signals[1] for pk in merged_peaks])
# build the ranks vector
idr.log("Ranking peaks", 'VERBOSE')
r1, r2 = build_rank_vectors(merged_peaks)
if args.only_merge_peaks:
localIDRs, IDRs = None, None
else:
if len(merged_peaks) < 20:
error_msg = "Peak files must contain at least 20 peaks post-merge"
error_msg += "\nHint: Merged peaks were written to the output file"
write_results_to_file(merged_peaks, args.output_file,
args.output_file_type, signal_type)
raise ValueError(error_msg)
localIDRs = fit_model_and_calc_local_idr(
r1,
r2,
starting_point=(args.initial_mu, args.initial_sigma,
args.initial_rho, args.initial_mix_param),
max_iter=args.max_iter,
convergence_eps=args.convergence_eps,
fix_mu=args.fix_mu,
fix_sigma=args.fix_sigma)
# if the use chose to use the best multi summit IDR, then
# make the correction and plot just the corrected peaks
if args.use_best_multisummit_IDR:
update_indices, localIDRs = correct_multi_summit_peak_IDR_values(
localIDRs, merged_peaks)
IDRs = calc_global_IDR(localIDRs)
if args.plot:
assert len(args.samples) == 2
plot(args, [s1[update_indices], s2[update_indices]],
[r1[update_indices], r2[update_indices]],
IDRs[update_indices],
args.output_file.name + ".noalternatesummitpeaks")
# we wrap this in an else statement to avoid calculating the global IDRs
# twice
else:
IDRs = calc_global_IDR(localIDRs)
if args.plot:
assert len(args.samples) == 2
plot(args, [s1, s2], [r1, r2], IDRs)
num_peaks_passing_thresh = write_results_to_file(
merged_peaks,
args.output_file,
args.output_file_type,
signal_type,
localIDRs=localIDRs,
IDRs=IDRs,
max_allowed_idr=args.idr_threshold,
soft_max_allowed_idr=args.soft_idr_threshold,
useBackwardsCompatibleOutput=args.use_old_output_format)
args.output_file.close()
