def setUpClass(cls):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/twold_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
(DIR + '/simulate_test/sumstats/' + str(i) for i in xrange(N_REP)))
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)
args.intercept_gencov = ','.join(('0' for _ in xrange(N_REP)))
args.intercept_h2 = ','.join(('1' for _ in xrange(N_REP)))
y = s.estimate_rg(args, log)
cls.rg = x
cls.rg_noint = y
def test_no_check_alleles(self):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)[0]
args.no_check_alleles = True
y = s.estimate_rg(args, log)[0]
assert_equal(x.rg_ratio, y.rg_ratio)
assert_almost_equal(x.rg_jknife, y.rg_jknife)
assert_equal(x.rg_se, y.rg_se)
def test_no_check_alleles(self):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)[0]
args.no_check_alleles = True
y = s.estimate_rg(args, log)[0]
assert_equal(x.rg_ratio, y.rg_ratio)
assert_almost_equal(x.rg_jknife, y.rg_jknife)
assert_equal(x.rg_se, y.rg_se)
def setUpClass(cls):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/twold_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
(DIR + '/simulate_test/sumstats/' + str(i) for i in xrange(N_REP)))
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)
args.intercept_gencov = ','.join(('0' for _ in xrange(N_REP)))
args.intercept_h2 = ','.join(('1' for _ in xrange(N_REP)))
y = s.estimate_rg(args, log)
cls.rg = x
cls.rg_noint = y
def test_twostep_rg(self):
# make sure two step isn't going crazy
args = parser.parse_args('')
args.ref_ld_chr = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/rg'
args.two_step = 999
x = s.estimate_rg(args, log)[0]
args.two_step = 99999
y = s.estimate_rg(args, log)[0]
assert_allclose(x.rg_ratio, y.rg_ratio, atol=1e-5)
assert_allclose(x.gencov.tot, y.gencov.tot, atol=1e-5)
def test_twostep_rg(self):
# make sure two step isn't going crazy
args = parser.parse_args('')
args.ref_ld_chr = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/rg'
args.two_step = 999
x = s.estimate_rg(args, log)[0]
args.two_step = 99999
y = s.estimate_rg(args, log)[0]
assert_allclose(x.rg_ratio, y.rg_ratio, atol=1e-5)
assert_allclose(x.gencov.tot, y.gencov.tot, atol=1e-5)
def test_rg_M(self):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)[0]
args.M = open(DIR + '/simulate_test/ldscore/oneld_onefile.l2.M_5_50',
'rb').read().rstrip('\n')
y = s.estimate_rg(args, log)[0]
assert_array_almost_equal(x.rg_ratio, y.rg_ratio)
assert_array_almost_equal(x.rg_se, y.rg_se)
args.M = '1,2'
assert_raises(ValueError, s.estimate_rg, args, log)
args.M = 'foo_bar'
assert_raises(ValueError, s.estimate_rg, args, log)
def test_rg_M(self):
args = parser.parse_args('')
args.ref_ld = DIR + '/simulate_test/ldscore/oneld_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
x = s.estimate_rg(args, log)[0]
args.M = open(
DIR + '/simulate_test/ldscore/oneld_onefile.l2.M_5_50', 'rb').read().rstrip('\n')
y = s.estimate_rg(args, log)[0]
assert_array_almost_equal(x.rg_ratio, y.rg_ratio)
assert_array_almost_equal(x.rg_se, y.rg_se)
args.M = '1,2'
assert_raises(ValueError, s.estimate_rg, args, log)
args.M = 'foo_bar'
assert_raises(ValueError, s.estimate_rg, args, log)
def test_rg_ref_ld(self):
args = parser.parse_args('')
args.ref_ld_chr = DIR + '/simulate_test/ldscore/twold_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
args.print_cov = True # right now just check no runtime errors
args.print_delete_vals = True
x = s.estimate_rg(args, log)[0]
args.ref_ld = DIR + '/simulate_test/ldscore/twold_firstfile,' + \
DIR + '/simulate_test/ldscore/twold_secondfile'
y = s.estimate_rg(args, log)[0]
args.ref_ld_chr = DIR + '/simulate_test/ldscore/twold_firstfile,' + \
DIR + '/simulate_test/ldscore/twold_secondfile'
z = s.estimate_rg(args, log)[0]
assert_almost_equal(x.rg_ratio, y.rg_ratio)
assert_almost_equal(y.rg_jknife, z.rg_jknife)
assert_almost_equal(x.rg_se, y.rg_se)
def test_rg_ref_ld(self):
args = parser.parse_args('')
args.ref_ld_chr = DIR + '/simulate_test/ldscore/twold_onefile'
args.w_ld = DIR + '/simulate_test/ldscore/w'
args.rg = ','.join(
[DIR + '/simulate_test/sumstats/1' for _ in xrange(2)])
args.out = DIR + '/simulate_test/1'
args.print_cov = True # right now just check no runtime errors
args.print_delete_vals = True
x = s.estimate_rg(args, log)[0]
args.ref_ld = DIR + '/simulate_test/ldscore/twold_firstfile,' + \
DIR + '/simulate_test/ldscore/twold_secondfile'
y = s.estimate_rg(args, log)[0]
args.ref_ld_chr = DIR + '/simulate_test/ldscore/twold_firstfile,' + \
DIR + '/simulate_test/ldscore/twold_secondfile'
z = s.estimate_rg(args, log)[0]
assert_almost_equal(x.rg_ratio, y.rg_ratio)
assert_almost_equal(y.rg_jknife, z.rg_jknife)
assert_almost_equal(x.rg_se, y.rg_se)
if not args.overlap_annot or args.not_M_5_50:
if args.frqfile is not None or args.frqfile_chr is not None:
log.log(
'The frequency file is unnecessary and is being ignored.'
)
args.frqfile = None
args.frqfile_chr = None
if args.overlap_annot and not args.not_M_5_50:
if not ((args.frqfile and args.ref_ld) or
(args.frqfile_chr and args.ref_ld_chr)):
raise ValueError(
'Must set either --frqfile and --ref-ld or --frqfile-chr and --ref-ld-chr'
)
if args.rg:
sumstats.estimate_rg(args, log)
elif args.h2:
sumstats.estimate_h2(args, log)
elif args.h2_cts:
sumstats.cell_type_specific(args, log)
# bad flags
else:
print header
print 'Error: no analysis selected.'
print 'ldsc.py -h describes options.'
except Exception:
ex_type, ex, tb = sys.exc_info()
log.log(traceback.format_exc(ex))
raise
finally:
def ldsc_rg_pair(args, **kwargs):
"""
Args is a list with elements:
- args[0] = phenotype name
- args[1] = phenotype description
- args[2] = file name for phenotype 1
- args[3] = file name for phenotype 2
- args[4] = N for phenotype 1
- args[5] = N_cases for phenotype 1
- args[6] = N_controls for phenotype 1
- args[7] = N for phenotype 2
- args[8] = N_cases for phenotype 2
- args[9] = N_controls for phenotype 2
Assumes keyword args for:
wd
gs_sumstat_dir
ld_ref_panel
"""
# handle args
phname = str(args[0])
phdesc = str(args[1])
f1 = str(args[2])
f2 = str(args[3])
n1 = int(args[4])
ncas1 = float(args[5])
ncon1 = float(args[6])
n2 = int(args[7])
ncas2 = float(args[8])
ncon2 = float(args[9])
# log
print "Starting phenotype: " + str(phname)
# download sumstats for phens
gs_ss_path1 = gs_sumstat_dir + '/' + str(f1)
loc_ss_path1 = wd + '/' + str(f1)
subprocess.call(['gsutil', 'cp', gs_ss_path1, loc_ss_path1])
gs_ss_path2 = gs_sumstat_dir + '/' + str(f2)
loc_ss_path2 = wd + '/' + str(f2)
subprocess.call(['gsutil', 'cp', gs_ss_path2, loc_ss_path2])
# list of files
rg_file_list = ','.join([loc_ss_path1, loc_ss_path2])
# list of names
rg_name_list = [str(f1), str(f2)]
# dummy output name
rg_out = wd + '/' + 'rg.summary'
# args for ldsc
args_ldsc_rg = Namespace(out=rg_out,
bfile=None,
l2=None,
extract=None,
keep=None,
ld_wind_snps=None,
ld_wind_kb=None,
ld_wind_cm=None,
print_snps=None,
annot=None,
thin_annot=False,
cts_bin=None,
cts_break=None,
cts_names=None,
per_allele=False,
pq_exp=None,
no_print_annot=False,
maf=None,
h2=None,
rg=rg_file_list,
ref_ld=None,
ref_ld_chr=ld_ref_panel,
w_ld=None,
w_ld_chr=ld_ref_panel,
overlap_annot=False,
no_intercept=False,
intercept_h2=None,
intercept_gencov=None,
M=None,
two_step=99999,
chisq_max=99999,
print_cov=False,
print_delete_vals=False,
chunk_size=50,
pickle=False,
invert_anyway=False,
yes_really=False,
n_blocks=200,
not_M_5_50=False,
return_silly_things=False,
no_check_alleles=False,
print_coefficients=False,
samp_prev=None,
pop_prev=None,
frqfile=None,
h2_cts=None,
frqfile_chr=None,
print_all_cts=False)
# run rg
rg_out = sumstats.estimate_rg(args_ldsc_rg, Logger_to_Logging())
# get basic rg summary table
rg_tab_txt = sumstats._get_rg_table(rg_name_list, rg_out, args_ldsc_rg)
rg_df = pd.read_csv(StringIO(rg_tab_txt), delim_whitespace=True)
print(rg_df)
# rename h2, int columns so we can add h2/int for phenotype 1
rg_df.rename(
{
'h2_obs': 'ph2_h2_obs',
'h2_obs_se': 'ph2_h2_obs_se',
'h2_int': 'ph2_h2_int',
'h2_int_se': 'ph2_h2_int_se'
},
axis='columns',
inplace=True)
# add h2/int for phenotype 1
t = lambda attr: lambda obj: getattr(obj, attr, 'NA')
rg_df['ph1_h2_int'] = map(t('intercept'), map(t('hsq1'), rg_out))
rg_df['ph1_h2_int_se'] = map(t('intercept_se'), map(t('hsq1'), rg_out))
rg_df['ph1_h2_obs'] = map(t('tot'), map(t('hsq1'), rg_out))
rg_df['ph1_h2_obs_se'] = map(t('tot_se'), map(t('hsq1'), rg_out))
# add phenotype info
rg_df.insert(0, 'description', str(phdesc))
rg_df.insert(0, 'phenotype', str(phname))
# add sample size info
rg_df['ph1_n'] = n1
rg_df['ph1_n_case'] = ncas1
rg_df['ph1_n_control'] = ncon1
rg_df['ph2_n'] = n2
rg_df['ph2_n_case'] = ncas2
rg_df['ph2_n_control'] = ncon2
print '#########'
print 'rg: ' + str(rg_df['rg'])
print '#########'
return rg_df
if args.w_ld and args.w_ld_chr:
raise ValueError('Cannot set both --w-ld and --w-ld-chr.')
if (args.samp_prev is not None) != (args.pop_prev is not None):
raise ValueError('Must set both or neither of --samp-prev and --pop-prev.')
if not args.overlap_annot or args.not_M_5_50:
if args.frqfile is not None or args.frqfile_chr is not None:
log.log('The frequency file is unnecessary and is being ignored.')
args.frqfile = None
args.frqfile_chr = None
if args.overlap_annot and not args.not_M_5_50:
if not ((args.frqfile and args.ref_ld) or (args.frqfile_chr and args.ref_ld_chr)):
raise ValueError ('Must set either --frqfile and --ref-ld or --frqfile-chr and --ref-ld-chr')
if args.rg:
sumstats.estimate_rg(args, log)
elif args.h2:
sumstats.estimate_h2(args, log)
# bad flags
else:
print header
print 'Error: no analysis selected.'
print 'ldsc.py -h describes options.'
except Exception:
ex_type, ex, tb = sys.exc_info()
log.log( traceback.format_exc(ex) )
raise
finally:
log.log('Analysis finished at {T}'.format(T=time.ctime()) )
time_elapsed = round(time.time()-start_time,2)
def ldsc_rg_target(ph_list, **kwargs):
"""
Assumes keyword args for:
wd
gs_sumstat_dir
ld_ref_panel
target_name
"""
# log
print "Starting phenotypes: "
print ph_list
# download sumstats for phens
for ph in ph_list:
gs_ss_path = gs_sumstat_dir + '/' + str(ph) + '.ukbb.sumstats.gz'
loc_ss_path = wd + '/' + str(ph) + '.ukbb.sumstats.gz'
subprocess.call(['gsutil', 'cp', gs_ss_path, loc_ss_path])
# list of files
ukb_loc_list = ','.join(
[wd + '/' + str(x) + ".ukbb.sumstats.gz" for x in ph_list])
rg_file_list = ','.join([loc_target_ss, ukb_loc_list])
# list of names
ukb_name_list = [str(x) + ".ukbb" for x in ph_list]
rg_name_list = [target_name] + ukb_name_list
# dummy output name
rg_out = wd + '/' + 'rg.summary'
# args for ldsc
args_ldsc_rg = Namespace(out=rg_out,
bfile=None,
l2=None,
extract=None,
keep=None,
ld_wind_snps=None,
ld_wind_kb=None,
ld_wind_cm=None,
print_snps=None,
annot=None,
thin_annot=False,
cts_bin=None,
cts_break=None,
cts_names=None,
per_allele=False,
pq_exp=None,
no_print_annot=False,
maf=None,
h2=None,
rg=rg_file_list,
ref_ld=None,
ref_ld_chr=ld_ref_panel,
w_ld=None,
w_ld_chr=ld_ref_panel,
overlap_annot=False,
no_intercept=False,
intercept_h2=None,
intercept_gencov=None,
M=None,
two_step=None,
chisq_max=None,
print_cov=False,
print_delete_vals=False,
chunk_size=50,
pickle=False,
invert_anyway=False,
yes_really=False,
n_blocks=200,
not_M_5_50=False,
return_silly_things=False,
no_check_alleles=False,
print_coefficients=False,
samp_prev=None,
pop_prev=None,
frqfile=None,
h2_cts=None,
frqfile_chr=None,
print_all_cts=False)
# run rg
rg_out = sumstats.estimate_rg(args_ldsc_rg, Logger_to_Logging())
# format output
rg_tab_txt = sumstats._get_rg_table(rg_name_list, rg_out, args_ldsc_rg)
rg_df = pd.read_csv(StringIO(rg_tab_txt), delim_whitespace=True)
return rg_df
