def main(p_dict):
bimfile = None
if p_dict['vbim'] is not None:
bimfile = p_dict['vbim']
elif p_dict['vgf'] is not None:
bimfile = p_dict['vgf'] + '.bim'
elif p_dict['gf'] is not None:
bimfile = p_dict['gf'] + '.bim'
else:
print('Set of validation SNPs is missing! Please specify either a validation PLINK genotype file, ' \
'or a PLINK BIM file with the SNPs of interest.')
if os.path.isfile(p_dict['out']):
print('Output file (%s) already exists! Delete, rename it, or use a different output file.'\
% (p_dict['out']))
raise Exception('Output file already exists!')
h5f = h5py.File(p_dict['out'], 'w')
summary_dict = {}
summary_dict[0]={'name':'Summary statistics filename:','value':p_dict['ssf']}
summary_dict[1]={'name':'LD reference genotypes filename:','value':p_dict['gf']}
summary_dict[3]={'name':'Coordinated data output filename:','value':p_dict['out']}
if p_dict['vgf'] is not None:
summary_dict[2]={'name':'Validation genotypes filename:','value':p_dict['vgf']}
sum_stats_parsers.parse_sum_stats(h5f, p_dict, bimfile, summary_dict)
coordinate_datasets(p_dict['gf'], h5f,summary_dict,
validation_genotype_file=p_dict['vgf'],
max_freq_discrep=p_dict['max_freq_discrep'],
min_maf=p_dict['maf'],
skip_coordination=p_dict['skip_coordination'],
debug=p_dict['debug'])
h5f.close()
reporting.print_summary(summary_dict, 'Summary of coordination step')
def main(p_dict):
#Check parameters
summary_dict = {}
summary_dict[0]={'name':'Coordinated data filename','value':p_dict['cf']}
summary_dict[0.1]={'name':'SNP weights output file (prefix)', 'value':p_dict['out']}
summary_dict[0.2]={'name':'LD data filename (prefix)', 'value':p_dict['ldf']}
summary_dict[1]={'name':'LD radius used','value':str(p_dict['ldr'])}
t0 = time.time()
summary_dict[1.09]={'name':'dash', 'value':'LD information'}
ld_dict = ld.get_ld_dict(p_dict['cf'], p_dict['ldf'], p_dict['ldr'], verbose=p_dict['debug'],
compressed=not p_dict['no_ld_compression'], use_hickle=p_dict['hickle_ld'], summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.2]={'name':'Running time for calculating LD information:','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
t0 = time.time()
summary_dict[1.9]={'name':'dash', 'value':'LDpred Gibbs sampler'}
ldpred_genomewide(data_file=p_dict['cf'], out_file_prefix=p_dict['out'], ps=p_dict['f'], ld_radius=p_dict['ldr'],
ld_dict=ld_dict, n=p_dict['N'], num_iter=p_dict['n_iter'], burn_in=p_dict['n_burn_in'],
h2=p_dict['h2'], use_gw_h2=p_dict['use_gw_h2'], verbose=p_dict['debug'], summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[2.2]={'name':'Running time for Gibbs sampler(s):','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
reporting.print_summary(summary_dict, 'Summary of LDpred Gibbs')
def main(p_dict):
bimfile = None
if p_dict['vbim'] is not None:
bimfile = p_dict['vbim']
elif p_dict['vgf'] is not None:
bimfile = p_dict['vgf'] + '.bim'
elif p_dict['gf'] is not None:
bimfile = p_dict['gf'] + '.bim'
else:
print('Set of validation SNPs is missing! Please specify either a validation PLINK genotype file, ' \
'or a PLINK BIM file with the SNPs of interest.')
if os.path.isfile(p_dict['out']):
print('Output file (%s) already exists! Delete, rename it, or use a different output file.'\
% (p_dict['out']))
raise Exception('Output file already exists!')
h5f = h5py.File(p_dict['out'], 'w')
summary_dict = {}
summary_dict[0]={'name':'Summary statistics filename:','value':p_dict['ssf']}
summary_dict[1]={'name':'LD reference genotypes filename:','value':p_dict['gf']}
summary_dict[3]={'name':'Coordinated data output filename:','value':p_dict['out']}
if p_dict['vgf'] is not None:
summary_dict[2]={'name':'Validation genotypes filename:','value':p_dict['vgf']}
sum_stats_parsers.parse_sum_stats(h5f, p_dict, bimfile, summary_dict)
coordinate_datasets(p_dict['gf'], h5f,summary_dict,
validation_genotype_file=p_dict['vgf'],
max_freq_discrep=p_dict['max_freq_discrep'],
min_maf=p_dict['maf'],
skip_coordination=p_dict['skip_coordination'],
debug=p_dict['debug'])
h5f.close()
reporting.print_summary(summary_dict, 'Summary of coordination step')
def main(p_dict):
#Check parameters
summary_dict = {}
summary_dict[0] = {
'name': 'Coordinated data filename',
'value': p_dict['cf']
}
summary_dict[0.1] = {
'name': 'SNP weights output file (prefix)',
'value': p_dict['out']
}
eff_type = get_eff_type(p_dict['cf'])
#If already BLUP betas, then skip LD calculation
if eff_type != 'BLUP':
summary_dict[0.2] = {
'name': 'LD data filename (prefix)',
'value': p_dict['ldf']
}
summary_dict[1.01] = {
'name': 'LD radius used',
'value': str(p_dict['ldr'])
}
summary_dict[1] = {'name': 'dash', 'value': 'LD information'}
t0 = time.time()
ld_dict = ld.get_ld_dict_using_p_dict(p_dict, summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.2] = {
'name': 'Running time for calculating LD information:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
t0 = time.time()
summary_dict[1.9] = {'name': 'dash', 'value': 'LDpred-fast'}
ldpred_fast_genomewide(
data_file=p_dict['cf'],
out_file_prefix=p_dict['out'],
ps=p_dict['f'],
ld_radius=p_dict['ldr'],
ld_dict=ld_dict,
n=p_dict['N'],
h2=p_dict['h2'],
use_gw_h2=p_dict['use_gw_h2'],
eff_type=eff_type,
summary_dict=summary_dict,
debug=p_dict['debug'],
)
t1 = time.time()
t = (t1 - t0)
summary_dict[3] = {
'name': 'Running time for LDpred-fast:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
reporting.print_summary(summary_dict, 'Summary of LDpred-fast')
def main(p_dict):
summary_dict = {}
summary_dict[0] = {
'name': 'Coordinated data filename',
'value': p_dict['cf']
}
summary_dict[0.1] = {
'name': 'SNP weights output file (prefix)',
'value': p_dict['out']
}
summary_dict[0.2] = {
'name': 'LD data filename (prefix)',
'value': p_dict['ldf']
}
summary_dict[1] = {'name': 'LD radius used', 'value': str(p_dict['ldr'])}
t0 = time.time()
summary_dict[1.09] = {'name': 'dash', 'value': 'LD information'}
ld_dict = ld.get_ld_dict_using_p_dict(p_dict, summary_dict={})
t1 = time.time()
t = (t1 - t0)
summary_dict[1.2] = {
'name': 'Running time for calculating LD information:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
t0 = time.time()
summary_dict[1.9] = {'name': 'dash', 'value': 'LDpred infinitesimal model'}
ldpred_inf_genomewide(data_file=p_dict['cf'],
out_file_prefix=p_dict['out'],
ld_radius=p_dict['ldr'],
ld_dict=ld_dict,
n=p_dict['N'],
h2=p_dict['h2'],
use_gw_h2=p_dict['use_gw_h2'],
verbose=p_dict['debug'],
summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[2.2] = {
'name': 'Running time for LDpred-inf:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
reporting.print_summary(summary_dict, 'Summary of LDpred-inf')
def main(p_dict):
summary_dict = {}
summary_dict[0]={'name':'Coordinated data filename','value':p_dict['cf']}
summary_dict[0.1]={'name':'SNP weights output file (prefix)', 'value':p_dict['out']}
summary_dict[0.2]={'name':'LD data filename (prefix)', 'value':p_dict['ldf']}
summary_dict[1]={'name':'LD radius used','value':str(p_dict['ldr'])}
t0 = time.time()
summary_dict[1.09]={'name':'dash', 'value':'LD information'}
ld_dict = ld.get_ld_dict(p_dict['cf'], p_dict['ldf'], p_dict['ldr'], verbose=p_dict['debug'],
compressed=not p_dict['no_ld_compression'], use_hickle=p_dict['hickle_ld'], summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.2]={'name':'Running time for calculating LD information:','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
t0 = time.time()
summary_dict[1.9]={'name':'dash', 'value':'LDpred Gibbs sampler'}
ldpred_genomewide(data_file=p_dict['cf'], out_file_prefix=p_dict['out'], ps=p_dict['f'], ld_radius=p_dict['ldr'],
ld_dict=ld_dict, n=p_dict['N'], num_iter=p_dict['n_iter'], burn_in=p_dict['n_burn_in'],
h2=p_dict['h2'], verbose=p_dict['debug'], summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[2.2]={'name':'Running time for Gibbs sampler(s):','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
reporting.print_summary(summary_dict, 'Summary of LDpred Gibbs')
def main(p_dict):
summary_dict = {}
summary_dict[0] = {
'name': 'Coordinated data filename',
'value': p_dict['cf']
}
summary_dict[0.1] = {
'name': 'SNP weights output file (prefix)',
'value': p_dict['out']
}
summary_dict[1] = {'name': 'LD radius used', 'value': str(p_dict['ldr'])}
t0 = time.time()
summary_dict[1.1] = {'name': 'dash', 'value': 'LD-pruning + Thresholding'}
run_pt(p_dict, summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[2] = {
'name': 'Running time for calculating P+T:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
reporting.print_summary(summary_dict,
'Summary of LD-pruning + Tresholding')
def main(p_dict):
summary_dict = {}
non_zero_chromosomes = set()
verbose = p_dict['debug']
t0 = time.time()
summary_dict[0] = {
'name': 'Validation genotype file (prefix):',
'value': p_dict['gf']
}
summary_dict[0.1] = {
'name': 'Input weight file(s) (prefix):',
'value': p_dict['rf']
}
summary_dict[0.2] = {
'name': 'Output scores file(s) (prefix):',
'value': p_dict['out']
}
adjust_for_pcs = False
adjust_for_covs = False
if not p_dict['only_score']:
summary_dict[0.9] = {'name': 'dash', 'value': 'Phenotypes'}
print('Parsing phenotypes')
if p_dict['pf'] is None:
if p_dict['gf'] is not None:
phen_map = parse_phen_file(p_dict['gf'] + '.fam',
'FAM',
verbose=verbose,
summary_dict=summary_dict)
else:
raise Exception('Validation phenotypes were not found.')
else:
phen_map = parse_phen_file(p_dict['pf'],
p_dict['pf_format'],
verbose=verbose,
summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.1] = {
'name': 'Individuals with phenotype information:',
'value': len(phen_map)
}
summary_dict[1.2] = {
'name': 'Running time for parsing phenotypes:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
if p_dict['cov_file'] != None:
adjust_for_covs = True
if verbose:
print('Parsing additional covariates')
with open(p_dict['cov_file'], 'r') as f:
num_missing = 0
for line in f:
l = line.split()
iid = l[0]
if iid in phen_map:
covariates = list(map(float, l[1:]))
phen_map[iid]['covariates'] = covariates
else:
num_missing += 1
if num_missing > 0:
summary_dict[2.1] = {
'name': 'Individuals w missing covariate information:',
'value': num_missing
}
if verbose:
print('Unable to find %d iids in phen file!' %
num_missing)
summary_dict[2] = {
'name': 'Parsed covariates file:',
'value': p_dict['cov_file']
}
if p_dict['pcs_file']:
adjust_for_pcs = True
if verbose:
print('Parsing PCs')
with open(p_dict['pcs_file'], 'r') as f:
num_missing = 0
for line in f:
l = line.split()
iid = l[1]
if iid in phen_map:
pcs = list(map(float, l[2:]))
phen_map[iid]['pcs'] = pcs
else:
num_missing += 1
if num_missing > 0:
summary_dict[3.1] = {
'name': 'Individuals w missing PCs:',
'value': num_missing
}
if verbose:
print('Unable to find %d iids in phen file!' %
num_missing)
summary_dict[3] = {
'name': 'Parsed PCs file:',
'value': p_dict['pcs_file']
}
num_individs = len(phen_map)
assert num_individs > 0, 'No phenotypes were found!'
else:
phen_map = None
t0 = time.time()
prs_file_is_missing = True
res_dict = {}
if p_dict['rf_format'] == 'LDPRED' or p_dict['rf_format'] == 'ANY':
weights_file = '%s_LDpred-inf.txt' % (p_dict['rf'])
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred-inf risk scores')
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred-inf.txt' % (p_dict['out'])
res_dict['LDpred_inf'] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if not p_dict['only_score']:
summary_dict[5.2] = {
'name': 'LDpred_inf (unadjusted) Pearson R2:',
'value': '%0.4f' % res_dict['LDpred_inf']['pred_r2']
}
prs_file_is_missing = False
best_ldpred_pred_r2 = 0
best_p = None
for p in p_dict['f']:
weights_file = '%s_LDpred_p%0.4e.txt' % (p_dict['rf'], p)
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred risk scores using f=%0.3e' % p)
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred_p%0.4e.txt' % (p_dict['out'], p)
method_str = 'LDpred_p%0.4e' % (p)
res_dict[method_str] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if len(res_dict[method_str]) and (
res_dict[method_str]['pred_r2']) > best_ldpred_pred_r2:
best_ldpred_pred_r2 = res_dict[method_str]['pred_r2']
best_p = p
prs_file_is_missing = False
if best_ldpred_pred_r2 > 0 and not p_dict['only_score']:
summary_dict[5.3] = {
'name': 'Best LDpred (f=%0.2e) (unadjusted) R2:' % (best_p),
'value': '%0.4f' % best_ldpred_pred_r2
}
# Plot results?
if p_dict['rf_format'] == 'P+T' or p_dict['rf_format'] == 'ANY':
best_pt_pred_r2 = 0
best_t = None
best_r2 = None
for max_r2 in p_dict['r2']:
for p_thres in p_dict['p']:
weights_file = '%s_P+T_r%0.2f_p%0.4e.txt' % (p_dict['rf'],
max_r2, p_thres)
if os.path.isfile(weights_file):
print(
'Calculating P+T risk scores using p-value threshold of %0.3e, and r2 threshold of %0.2f'
% (p_thres, max_r2))
rs_id_map, non_zero_chromosomes = parse_pt_res(
weights_file)
if len(rs_id_map) > 0:
out_file = '%s_P+T_p%0.4e.txt' % (p_dict['out'],
p_thres)
method_str = 'P+T_p%0.4e' % (p_thres)
res_dict[method_str] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
non_zero_chromosomes=non_zero_chromosomes,
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if len(res_dict[method_str]) and (res_dict[method_str][
'pred_r2']) > best_pt_pred_r2:
best_pt_pred_r2 = res_dict[method_str]['pred_r2']
best_t = p_thres
best_r2 = max_r2
else:
print(
'No SNPs found with p-values below the given threshold.'
)
prs_file_is_missing = False
if best_pt_pred_r2 > 0 and not p_dict['only_score']:
summary_dict[5.4] = {
'name':
'Best P+T (r2=%0.2f, p=%0.2e) (unadjusted) R2:' %
(best_r2, best_t),
'value':
'%0.4f' % best_pt_pred_r2
}
# Plot results?
assert not prs_file_is_missing, 'No SNP weights file was found. A prefix to these should be provided via the --rf flag. Note that the prefix should exclude the _LDpred_.. extension or file ending. '
#Identifying the best prediction
if not p_dict['only_score']:
best_pred_r2 = 0
best_method_str = None
for method_str in res_dict:
if len(res_dict[method_str]) and (
res_dict[method_str]['pred_r2']) > best_pred_r2:
best_pred_r2 = res_dict[method_str]['pred_r2']
best_method_str = method_str
if best_method_str is not None:
print(
'The highest (unadjusted) Pearson R2 was %0.4f, and provided by %s'
% (best_pred_r2, best_method_str))
summary_dict[5.99] = {
'name': 'dash',
'value': 'Optimal polygenic score'
}
summary_dict[6] = {
'name': 'Method with highest (unadjusted) Pearson R2:',
'value': best_method_str
}
summary_dict[6.1] = {
'name': 'Best (unadjusted) Pearson R2:',
'value': '%0.4f' % best_pred_r2
}
t1 = time.time()
t = (t1 - t0)
summary_dict[4.9] = {'name': 'dash', 'value': 'Scoring'}
summary_dict[5.9] = {
'name': 'Running time for calculating scores:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
if prs_file_is_missing:
print(
'SNP weights files were not found. This could be due to a mis-specified --rf flag, or other issues.'
)
reporting.print_summary(summary_dict, 'Scoring Summary')
def main(p_dict):
summary_dict = {}
non_zero_chromosomes = set()
verbose = p_dict['debug']
t0 = time.time()
summary_dict[0]={'name':'Validation genotype file (prefix):','value':p_dict['gf']}
summary_dict[0.1]={'name':'Input weight file(s) (prefix):','value':p_dict['rf']}
summary_dict[0.2]={'name':'Output scores file(s) (prefix):','value':p_dict['out']}
adjust_for_pcs=False
adjust_for_covs=False
if not p_dict['only_score']:
summary_dict[0.9]={'name':'dash', 'value':'Phenotypes'}
print('Parsing phenotypes')
if p_dict['pf'] is None:
if p_dict['gf'] is not None:
phen_map = parse_phen_file(p_dict['gf'] + '.fam', 'FAM', verbose=verbose, summary_dict=summary_dict)
else:
raise Exception('Validation phenotypes were not found.')
else:
phen_map = parse_phen_file(p_dict['pf'], p_dict['pf_format'], verbose=verbose, summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.1]={'name':'Individuals with phenotype information:','value':len(phen_map)}
summary_dict[1.2]={'name':'Running time for parsing phenotypes:','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
if p_dict['cov_file'] != None:
adjust_for_covs=True
if verbose:
print('Parsing additional covariates')
with open(p_dict['cov_file'], 'r') as f:
num_missing = 0
for line in f:
l = line.split()
iid = l[0]
if iid in phen_map:
covariates = list(map(float, l[1:]))
phen_map[iid]['covariates'] = covariates
else:
num_missing += 1
if num_missing > 0:
summary_dict[2.1]={'name':'Individuals w missing covariate information:','value':num_missing}
if verbose:
print('Unable to find %d iids in phen file!' % num_missing)
summary_dict[2]={'name':'Parsed covariates file:','value':p_dict['cov_file']}
if p_dict['pcs_file']:
adjust_for_pcs=True
if verbose:
print('Parsing PCs')
with open(p_dict['pcs_file'], 'r') as f:
num_missing = 0
for line in f:
l = line.split()
iid = l[1]
if iid in phen_map:
pcs = list(map(float, l[2:]))
phen_map[iid]['pcs'] = pcs
else:
num_missing += 1
if num_missing > 0:
summary_dict[3.1]={'name':'Individuals w missing PCs:','value':num_missing}
if verbose:
print('Unable to find %d iids in phen file!' % num_missing)
summary_dict[3]={'name':'Parsed PCs file:','value':p_dict['pcs_file']}
num_individs = len(phen_map)
assert num_individs > 0, 'No phenotypes were found!'
else:
phen_map = None
t0 = time.time()
prs_file_is_missing = True
res_dict = {}
if p_dict['rf_format'] == 'LDPRED' or p_dict['rf_format']=='ANY':
weights_file = '%s_LDpred-inf.txt' % (p_dict['rf'])
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred-inf risk scores')
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred-inf.txt' % (p_dict['out'])
res_dict['LDpred_inf'] = calc_risk_scores(p_dict['gf'], rs_id_map, phen_map, out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose, summary_dict=summary_dict)
if not p_dict['only_score']:
summary_dict[5.2]={'name':'LDpred_inf (unadjusted) Pearson R2:','value':'%0.4f'%res_dict['LDpred_inf']['pred_r2']}
prs_file_is_missing = False
best_ldpred_pred_r2 = 0
best_p = None
for p in p_dict['f']:
weights_file = '%s_LDpred_p%0.4e.txt' % (p_dict['rf'], p)
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred risk scores using f=%0.3e' % p)
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred_p%0.4e.txt' % (p_dict['out'], p)
method_str = 'LDpred_p%0.4e' % (p)
res_dict[method_str] = calc_risk_scores(p_dict['gf'], rs_id_map, phen_map, out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose, summary_dict=summary_dict)
if len(res_dict[method_str]) and (res_dict[method_str]['pred_r2']) >best_ldpred_pred_r2:
best_ldpred_pred_r2 = res_dict[method_str]['pred_r2']
best_p = p
prs_file_is_missing=False
if best_ldpred_pred_r2>0 and not p_dict['only_score']:
summary_dict[5.3]={'name':'Best LDpred (f=%0.2e) (unadjusted) R2:'%(best_p),'value':'%0.4f'%best_ldpred_pred_r2}
# Plot results?
if p_dict['rf_format'] == 'P+T' or p_dict['rf_format']=='ANY':
best_pt_pred_r2 = 0
best_t = None
best_r2 = None
for max_r2 in p_dict['r2']:
for p_thres in p_dict['p']:
weights_file = '%s_P+T_r%0.2f_p%0.4e.txt' % (p_dict['rf'], max_r2, p_thres)
if os.path.isfile(weights_file):
print('')
print('Calculating P+T risk scores using p-value threshold of %0.3e, and r2 threshold of %0.2f' % (p_thres, max_r2))
rs_id_map, non_zero_chromosomes = parse_pt_res(weights_file)
out_file = '%s_P+T_p%0.4e.txt' % (p_dict['out'], p_thres)
method_str = 'P+T_p%0.4e' % (p_thres)
res_dict[method_str] = calc_risk_scores(p_dict['gf'], rs_id_map, phen_map, out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
non_zero_chromosomes=non_zero_chromosomes,
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose, summary_dict=summary_dict)
if len(res_dict[method_str]) and (res_dict[method_str]['pred_r2']) >best_pt_pred_r2:
best_pt_pred_r2 = res_dict[method_str]['pred_r2']
best_t = p_thres
best_r2 = max_r2
prs_file_is_missing=False
if best_pt_pred_r2>0 and not p_dict['only_score']:
summary_dict[5.4]={'name':'Best P+T (r2=%0.2f, p=%0.2e) (unadjusted) R2:'%(best_r2, best_t),'value':'%0.4f'%best_pt_pred_r2}
# Plot results?
assert not prs_file_is_missing, 'No SNP weights file was found. A prefix to these should be provided via the --rf flag. Note that the prefix should exclude the _LDpred_.. extension or file ending. '
#Identifying the best prediction
if not p_dict['only_score']:
best_pred_r2 = 0
best_method_str = None
for method_str in res_dict:
if len(res_dict[method_str]) and (res_dict[method_str]['pred_r2']) >best_pred_r2:
best_pred_r2 = res_dict[method_str]['pred_r2']
best_method_str = method_str
if best_method_str is not None:
print('The highest (unadjusted) Pearson R2 was %0.4f, and provided by %s'%(best_pred_r2,best_method_str))
summary_dict[5.99]={'name':'dash', 'value':'Optimal polygenic score'}
summary_dict[6]={'name':'Method with highest (unadjusted) Pearson R2:','value':best_method_str}
summary_dict[6.1]={'name':'Best (unadjusted) Pearson R2:','value':'%0.4f'%best_pred_r2}
t1 = time.time()
t = (t1 - t0)
summary_dict[4.9]={'name':'dash', 'value':'Scoring'}
summary_dict[5.9]={'name':'Running time for calculating scores:','value':'%d min and %0.2f secs'% (t / 60, t % 60)}
if prs_file_is_missing:
print('SNP weights files were not found. This could be due to a mis-specified --rf flag, or other issues.')
reporting.print_summary(summary_dict,'Scoring Summary')
def main(p_dict):
assert p_dict['summary_file'] is None or not p_dict[
'only_score'], 'Prediction summary file cannot be produced when the --only-score flag is set.'
summary_dict = {}
non_zero_chromosomes = set()
verbose = p_dict['debug']
t0 = time.time()
summary_dict[0] = {
'name': 'Validation genotype file (prefix):',
'value': p_dict['gf']
}
summary_dict[0.1] = {
'name': 'Input weight file(s) (prefix):',
'value': p_dict['rf']
}
summary_dict[0.2] = {
'name': 'Output scores file(s) (prefix):',
'value': p_dict['out']
}
adjust_for_pcs = False
adjust_for_covs = False
if not p_dict['only_score']:
summary_dict[0.9] = {'name': 'dash', 'value': 'Phenotypes'}
if verbose:
print('Parsing phenotypes')
if p_dict['pf'] is None:
if p_dict['gf'] is not None:
phen_map = parse_phen_file(p_dict['gf'] + '.fam',
'FAM',
verbose=verbose,
summary_dict=summary_dict)
else:
raise Exception('Validation phenotypes were not found.')
else:
phen_map = parse_phen_file(p_dict['pf'],
p_dict['pf_format'],
verbose=verbose,
summary_dict=summary_dict)
t1 = time.time()
t = (t1 - t0)
summary_dict[1.1] = {
'name': 'Individuals with phenotype information:',
'value': len(phen_map)
}
summary_dict[1.2] = {
'name': 'Running time for parsing phenotypes:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
if p_dict['cov_file'] != None:
adjust_for_covs = True
if verbose:
print('Parsing additional covariates')
parse_covariates(p_dict, phen_map, summary_dict, verbose)
if p_dict['pcs_file']:
adjust_for_pcs = True
if verbose:
print('Parsing PCs')
parse_pcs(p_dict, phen_map, summary_dict, verbose)
num_individs = len(phen_map)
assert num_individs > 0, 'No phenotypes were found!'
else:
phen_map = None
t0 = time.time()
prs_file_is_missing = True
res_dict = {}
if p_dict['rf_format'] == 'LDPRED' or p_dict['rf_format'] == 'ANY':
weights_file = '%s_LDpred-inf.txt' % (p_dict['rf'])
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred-inf risk scores')
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred-inf.txt' % (p_dict['out'])
res_dict['LDpred_inf'] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if not p_dict['only_score']:
summary_dict[5.2] = {
'name': 'LDpred_inf (unadjusted) Pearson R2:',
'value': '%0.4f' % res_dict['LDpred_inf']['pred_r2']
}
prs_file_is_missing = False
best_ldpred_pred_r2 = 0
best_p = None
for p in p_dict['f']:
weights_file = '%s_LDpred_p%0.4e.txt' % (p_dict['rf'], p)
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred risk scores using f=%0.3e' % p)
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred_p%0.4e.txt' % (p_dict['out'], p)
method_str = 'LDpred_p%0.4e' % (p)
res_dict[method_str] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if len(res_dict[method_str]) and (
res_dict[method_str]['pred_r2']) > best_ldpred_pred_r2:
best_ldpred_pred_r2 = res_dict[method_str]['pred_r2']
best_p = p
prs_file_is_missing = False
if best_ldpred_pred_r2 > 0 and not p_dict['only_score']:
summary_dict[5.3] = {
'name': 'Best LDpred (f=%0.2e) (unadjusted) R2:' % (best_p),
'value': '%0.4f' % best_ldpred_pred_r2
}
best_ldpred_fast_pred_r2 = 0
best_p = None
for p in p_dict['f']:
weights_file = '%s_LDpred_fast_p%0.4e.txt' % (p_dict['rf'], p)
if os.path.isfile(weights_file):
print('')
print('Calculating LDpred-fast risk scores using f=%0.3e' % p)
rs_id_map = parse_ldpred_res(weights_file)
out_file = '%s_LDpred_fast_p%0.4e.txt' % (p_dict['out'], p)
method_str = 'LDpred_fast_p%0.4e' % (p)
res_dict[method_str] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if len(res_dict[method_str]) and (res_dict[method_str][
'pred_r2']) > best_ldpred_fast_pred_r2:
best_ldpred_fast_pred_r2 = res_dict[method_str]['pred_r2']
best_p = p
prs_file_is_missing = False
if best_ldpred_fast_pred_r2 > 0 and not p_dict['only_score']:
summary_dict[5.4] = {
'name':
'Best LDpred-fast (f=%0.2e) (unadjusted) R2:' % (best_p),
'value': '%0.4f' % best_ldpred_fast_pred_r2
}
# Plot results?
if p_dict['rf_format'] == 'P+T' or p_dict['rf_format'] == 'ANY':
best_pt_pred_r2 = 0
best_t = None
best_r2 = None
for max_r2 in p_dict['r2']:
for p_thres in p_dict['p']:
weights_file = '%s_P+T_r%0.2f_p%0.4e.txt' % (p_dict['rf'],
max_r2, p_thres)
if os.path.isfile(weights_file):
print(
'Calculating P+T risk scores using p-value threshold of %0.3e, and r2 threshold of %0.2f'
% (p_thres, max_r2))
rs_id_map, non_zero_chromosomes = parse_pt_res(
weights_file)
if len(rs_id_map) > 0:
out_file = '%s_P+T_r%0.2f_p%0.4e.txt' % (
p_dict['out'], max_r2, p_thres)
method_str = 'P+T_r%0.2f_p%0.4e' % (max_r2, p_thres)
res_dict[method_str] = calc_risk_scores(
p_dict['gf'],
rs_id_map,
phen_map,
out_file=out_file,
split_by_chrom=p_dict['split_by_chrom'],
non_zero_chromosomes=non_zero_chromosomes,
adjust_for_pcs=adjust_for_pcs,
adjust_for_covariates=adjust_for_covs,
only_score=p_dict['only_score'],
verbose=verbose,
summary_dict=summary_dict)
if len(res_dict[method_str]) and (res_dict[method_str][
'pred_r2']) > best_pt_pred_r2:
best_pt_pred_r2 = res_dict[method_str]['pred_r2']
best_t = p_thres
best_r2 = max_r2
else:
print(
'No SNPs found with p-values below the given threshold.'
)
prs_file_is_missing = False
if best_pt_pred_r2 > 0 and not p_dict['only_score']:
summary_dict[5.5] = {
'name':
'Best P+T (r2=%0.2f, p=%0.2e) (unadjusted) R2:' %
(best_r2, best_t),
'value':
'%0.4f' % best_pt_pred_r2
}
# Plot results?
assert not prs_file_is_missing, 'No SNP weights file was found. A prefix to these should be provided via the --rf flag. Note that the prefix should exclude the _LDpred_.. extension or file ending. '
res_summary_file = p_dict['summary_file']
if res_summary_file is not None and not p_dict['only_score']:
with open(res_summary_file, 'w') as f:
if verbose:
print('Writing Results Summary to file %s' % res_summary_file)
out_str = 'Pred_Method Pred_corr Pred_R2 SNPs_used\n'
f.write(out_str)
for method_str in sorted(res_dict):
out_str = '%s %0.4f %0.4f %i\n' % (
method_str, res_dict[method_str]['corr_r2'],
res_dict[method_str]['pred_r2'],
res_dict[method_str]['num_snps'])
f.write(out_str)
#Identifying the best prediction
if not p_dict['only_score']:
best_pred_r2 = 0
best_method_str = None
for method_str in res_dict:
if len(res_dict[method_str]) and (
res_dict[method_str]['pred_r2']) > best_pred_r2:
best_pred_r2 = res_dict[method_str]['pred_r2']
best_method_str = method_str
if best_method_str is not None:
print(
'The highest (unadjusted) Pearson R2 was %0.4f, and provided by %s'
% (best_pred_r2, best_method_str))
summary_dict[5.99] = {
'name': 'dash',
'value': 'Optimal polygenic score'
}
summary_dict[6] = {
'name': 'Method with highest (unadjusted) Pearson R2:',
'value': best_method_str
}
summary_dict[6.1] = {
'name': 'Best (unadjusted) Pearson R2:',
'value': '%0.4f' % best_pred_r2
}
if verbose:
summary_dict[6.2] = {
'name': 'Number of SNPs used',
'value': '%d' % res_dict[best_method_str]['num_snps']
}
summary_dict[6.3] = {
'name': 'Number of SNPs flipped',
'value':
'%d' % res_dict[best_method_str]['num_flipped_nts']
}
summary_dict[6.4] = {
'name': 'Fraction of SNPs not found in validation data',
'value':
'%0.4f' % res_dict[best_method_str]['perc_missing']
}
summary_dict[6.5] = {
'name': 'Number of duplicated SNPs',
'value':
'%d' % res_dict[best_method_str]['duplicated_snps']
}
summary_dict[6.6] = {
'name':
'Number of non-matching nucleotides SNPs',
'value':
'%d' % res_dict[best_method_str]['num_non_matching_nts']
}
t1 = time.time()
t = (t1 - t0)
summary_dict[4.9] = {'name': 'dash', 'value': 'Scoring'}
summary_dict[5.9] = {
'name': 'Running time for calculating scores:',
'value': '%d min and %0.2f secs' % (t / 60, t % 60)
}
if prs_file_is_missing:
print(
'SNP weights files were not found. This could be due to a mis-specified --rf flag, or other issues.'
)
reporting.print_summary(summary_dict, 'Scoring Summary')
