def main(args):
# Set up parameters.
alpha = 0.01
n_itr = 1500
# Set up the output folder.
output_folder = os.path.realpath(
'..') + '/results/result_' + args.output_folder
output_datafile = '/data3/martin/gtex_data/results/result_' + args.output_folder + '.pickle'
if not os.path.exists(output_folder):
os.makedirs(output_folder)
else:
filelist = [os.remove(os.path.join(output_folder, f))\
for f in os.listdir(output_folder)]
# Load the data.
p, x, n_full, cate_name, cis_name = dl.load_GTEx(args.data_name)
# Logger.
logging.basicConfig(level=logging.INFO,format='%(module)s:: %(message)s',\
filename=output_folder+'/result.log', filemode='w')
logger = logging.getLogger()
result_dic = {}
# An overview of the data
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
# Report the baseline methods.
n_rej, t_rej = md.bh_test(p, alpha=alpha, n_full=n_full, verbose=False)
logger.info('## BH, n_rej=%d, t_rej=%0.5f' % (n_rej, t_rej))
result_dic['bh'] = {'h_hat': p < t_rej}
n_rej, t_rej, pi0_hat = md.sbh_test(p,
alpha=alpha,
n_full=n_full,
verbose=False)
result_dic['sbh'] = {'h_hat': p < t_rej}
logger.info('## SBH, n_rej=%d, t_rej=%0.5f, pi0_hat=%0.3f' %
(n_rej, t_rej, pi0_hat))
# Analysis
md.adafdr_explore(p, x, alpha=alpha, n_full=n_full, vis_dim=None, cate_name=cate_name,\
output_folder=output_folder, h=None)
# Fast mode.
output_folder_fast = output_folder + '_fast'
if not os.path.exists(output_folder_fast):
os.makedirs(output_folder_fast)
else:
filelist = [os.remove(os.path.join(output_folder_fast, f))\
for f in os.listdir(output_folder_fast)]
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p, x, K=5, alpha=alpha, h=None, n_full=n_full, n_itr=n_itr,\
verbose=True, output_folder=output_folder_fast, random_state=0,\
fast_mode=True)
n_rej = res['n_rej']
t_rej = res['threshold']
result_dic['nfdr (fast)'] = {'h_hat': p < t_rej}
logger.info('## nfdr2 (fast mode), n_rej1=%d, n_rej2=%d, n_rej_total=%d' %
(n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time (fast mode): %0.1fs' %
(time.time() - start_time))
# Full mode.
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p, x, K=5, alpha=alpha, h=None, n_full=n_full, n_itr=n_itr,\
verbose=True, output_folder=output_folder, random_state=0,\
fast_mode=False, single_core=False)
n_rej = res['n_rej']
t_rej = res['threshold']
result_dic['nfdr'] = {'h_hat': p < t_rej}
logger.info('## nfdr2, n_rej1=%d, n_rej2=%d, n_rej_total=%d' %
(n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time: %0.1fs' % (time.time() - start_time))
# Store the result
fil = open(output_datafile, 'wb')
pickle.dump(result_dic, fil)
fil.close()
def main(args):
# Set up parameters.
alpha = 0.1
n_itr = 1500
# Set up the output folder.
output_folder = os.path.realpath(
'..') + '/result_small_data/result_' + args.output_folder
if not os.path.exists(output_folder):
os.makedirs(output_folder)
else:
filelist = [os.remove(os.path.join(output_folder, f))\
for f in os.listdir(output_folder)]
# Load the data.
if 'GTEx' in args.data_loader:
p, x, n_full, cate_name = dl.load_GTEx_full(verbose=True)
x = x[:, 0:3]
h = None
else:
p, h, x = eval('dl.' + args.data_loader + '()')
n_full = p.shape[0]
cate_name = {}
# Logger.
logging.basicConfig(level=logging.INFO,format='%(module)s:: %(message)s',\
filename=output_folder+'/result.log', filemode='w')
logger = logging.getLogger()
# An overview of the data
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
# Report the baseline methods.
n_rej, t_rej = md.bh_test(p, alpha=alpha, n_full=n_full, verbose=False)
logger.info('## BH, n_rej=%d, t_rej=%0.5f' % (n_rej, t_rej))
n_rej, t_rej, pi0_hat = md.sbh_test(p,
alpha=alpha,
n_full=n_full,
verbose=False)
logger.info('## SBH, n_rej=%d, t_rej=%0.5f, pi0_hat=%0.3f' %
(n_rej, t_rej, pi0_hat))
# Analysis
md.adafdr_explore(p, x, alpha=alpha, n_full=n_full, vis_dim=None, cate_name=cate_name,\
output_folder=output_folder, h=None)
# Fast mode.
output_folder_fast = output_folder + '_fast'
if not os.path.exists(output_folder_fast):
os.makedirs(output_folder_fast)
else:
filelist = [os.remove(os.path.join(output_folder_fast, f))\
for f in os.listdir(output_folder_fast)]
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p, x, K=5, alpha=alpha, h=None, n_full=n_full, n_itr=n_itr,\
verbose=True, output_folder=output_folder_fast, random_state=0,\
fast_mode=True)
n_rej = res['n_rej']
t_rej = res['threshold']
logger.info('## AdaFDR (fast mode), n_rej1=%d, n_rej2=%d, n_rej_total=%d' %
(n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time (fast mode): %0.1fs' %
(time.time() - start_time))
# Full mode.
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p, x, K=5, alpha=alpha, h=None, n_full=n_full, n_itr=n_itr,\
verbose=True, output_folder=output_folder, random_state=0,\
fast_mode=False, single_core=False)
n_rej = res['n_rej']
t_rej = res['threshold']
logger.info('## AdaFDR, n_rej1=%d, n_rej2=%d, n_rej_total=%d' %
(n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time: %0.1fs' % (time.time() - start_time))
def main(args):
# Set up parameters.
alpha = 0.1
n_itr = 1500
# Set up the output folder.
output_folder = os.path.realpath('..') + '/result_simulation/result_' + args.output_folder
if not os.path.exists(output_folder):
os.makedirs(output_folder)
else:
filelist = [os.remove(os.path.join(output_folder, f))\
for f in os.listdir(output_folder)]
# Get logger.
logging.basicConfig(level=logging.INFO,format='%(message)s',\
filename=output_folder+'/result.log', filemode='w')
logger = logging.getLogger()
# Run method in all data in the folder
file_list = os.listdir(args.input_folder)
alpha_list = [0.05, 0.1, 0.15, 0.2]
if 'speed' in args.input_folder:
alpha_list = [0.1]
if 'ntest' in args.input_folder:
alpha_list = [0.1]
if 'prop_alt' in args.input_folder:
alpha_list = [0.1]
result_dic = {'bh': {}, 'sbh': {}, 'nfdr (fast)': {}, 'nfdr': {}}
time_dic = {'nfdr (fast)': {}, 'nfdr': {}}
for alpha in alpha_list:
result_dic['bh'][alpha] = []
result_dic['sbh'][alpha] = []
result_dic['nfdr (fast)'][alpha] = []
result_dic['nfdr'][alpha] = []
time_dic['nfdr (fast)'][alpha] = {}
time_dic['nfdr'][alpha] = {}
for filename in file_list:
filename_short = filename
if filename[0] == '.':
continue
print('# Processing %s with alpha=%0.2f'%(filename, alpha))
logger.info('# Processing %s with alpha=%0.2f'%(filename, alpha))
filename = args.input_folder + '/' + filename
p, x, h = dl.load_simulation_data(filename)
n_full = p.shape[0]
# Report the baseline.
n_rej, t_rej = md.bh_test(p, alpha=alpha, n_full=n_full, verbose=False)
result_dic['bh'][alpha].append([h, p<=t_rej])
logger.info('## BH, n_rej=%d, t_rej=%0.5f'%(n_rej,t_rej))
n_rej, t_rej, pi0_hat = md.sbh_test(p, alpha=alpha, n_full=n_full, verbose=False)
result_dic['sbh'][alpha].append([h, p<=t_rej])
logger.info('## SBH, n_rej=%d, t_rej=%0.5f, pi0_hat=%0.3f'%(n_rej, t_rej, pi0_hat))
# Fast mode.
start_time = time.time()
# res = md.adafdr_test(p, x, K=5, alpha=alpha, h=h, n_full=n_full, n_itr=n_itr,\
# verbose=False, output_folder=None, random_state=0,\
# fast_mode=True)
res_train = md.adafdr_train(p, x, K=5, alpha=alpha, h=h, n_full=n_full,
n_itr=n_itr, verbose=False,
output_folder=None, random_state=0,
fast_mode=True)
res = md.adafdr_test(res_train, alpha, n_full=n_full,
output_folder = None)
n_rej = res['n_rej']
t_rej = res['threshold']
time_dic['nfdr (fast)'][alpha][filename_short] = time.time()-start_time
result_dic['nfdr (fast)'][alpha].append([h, p<=t_rej])
logger.info('## nfdr2 (fast mode), n_rej1=%d, n_rej2=%d, n_rej_total=%d'%(n_rej[0],n_rej[1],n_rej[0]+n_rej[1]))
logger.info('## Total time (fast mode): %0.1fs'%(time.time()-start_time))
# Full mode.
start_time = time.time()
# res = md.adafdr_test(p, x, K=5, alpha=alpha, h=h, n_full=n_full, n_itr=n_itr,\
# verbose=False, output_folder=None, random_state=0,\
# fast_mode=False, single_core=False)
res_train = md.adafdr_train(p, x, K=5, alpha=alpha, h=h, n_full=n_full,
n_itr=n_itr, verbose=False,
output_folder=None, random_state=0,
fast_mode=False, single_core=False)
res = md.adafdr_test(res_train, alpha, n_full=n_full,
output_folder = None)
n_rej = res['n_rej']
t_rej = res['threshold']
time_dic['nfdr'][alpha][filename_short] = time.time()-start_time
result_dic['nfdr'][alpha].append([h, p<=t_rej])
logger.info('## nfdr2, n_rej1=%d, n_rej2=%d, n_rej_total=%d'%(n_rej[0],n_rej[1],n_rej[0]+n_rej[1]))
logger.info('## Total time: %0.1fs'%(time.time()-start_time))
logger.info('\n')
# Store the result
fil = open(output_folder+'/result_dic.pickle','wb')
pickle.dump(result_dic, fil)
pickle.dump(time_dic, fil)
fil.close()
def main(args):
# Set up the parameters.
input_folder = args.input_folder
output_folder = './temp_result/res_' + args.data_name
if args.alpha is not None:
alpha_list = [args.alpha]
else:
alpha_list = [0.05, 0.1, 0.15, 0.2]
if not os.path.exists(output_folder):
os.makedirs(output_folder)
else:
filelist = [os.remove(os.path.join(output_folder, f))\
for f in os.listdir(output_folder)]
print('input_folder: %s'%input_folder)
print('output_folder: %s'%output_folder)
print('alpha_list: %s'%alpha_list)
# Get a file for recording.
f_write = open(output_folder+'/result.log', 'w')
# Process all data in the folder
file_list = os.listdir(args.input_folder)
result_dic = {'bh': [], 'sbh': [], 'adafdr-fast': [], 'adafdr': []}
for filename in file_list:
if filename[0] == '.':
continue
file_path = args.input_folder + '/' + filename
p, x, h = dl.load_simulation_data(file_path)
for alpha in alpha_list:
print('# Processing %s with alpha=%0.2f'%(filename, alpha))
f_write.write('# Processing %s with alpha=%0.2f\n'%(filename, alpha))
# BH result
n_rej, t_rej = md.bh_test(p, alpha=alpha, verbose=False)
fdp,power = get_fdp_and_power(h, p<=t_rej)
result_dic['bh'].append([fdp, power, alpha, filename])
f_write.write('## BH discoveries: %d, threshold=%0.3f\n'%(n_rej,t_rej))
# SBH result
n_rej, t_rej, pi0_hat = md.sbh_test(p, alpha=alpha, verbose=False)
fdp,power = get_fdp_and_power(h, p<=t_rej)
result_dic['sbh'].append([fdp, power, alpha, filename])
temp = '## SBH discoveries: %d, threshold=%0.3f, pi0_hat=%0.3f\n'%(n_rej, t_rej, pi0_hat)
f_write.write(temp)
# AdaFDR-fast result
start_time = time.time()
res = md.adafdr_test(p, x, alpha=alpha, fast_mode=True)
n_rej = res['n_rej']
t_rej = res['threshold']
fdp,power = get_fdp_and_power(h, p<=t_rej)
result_dic['adafdr-fast'].append([fdp, power, alpha, filename])
temp = '## AdaFDR-fast discoveries: fold_1=%d, fold_2=%d, total=%d\n'%\
(n_rej[0],n_rej[1],n_rej[0]+n_rej[1])
f_write.write(temp)
f_write.write('## Time: %0.1fs'%(time.time()-start_time))
# AdaFDR result
start_time = time.time()
res = md.adafdr_test(p, x, alpha=alpha, fast_mode=False)
n_rej = res['n_rej']
t_rej = res['threshold']
fdp,power = get_fdp_and_power(h, p<=t_rej)
result_dic['adafdr'].append([fdp, power, alpha, filename])
temp = '## AdaFDR discoveries: fold_1=%d, fold_2=%d, total=%d\n'%\
(n_rej[0],n_rej[1],n_rej[0]+n_rej[1])
f_write.write(temp)
f_write.write('## Time: %0.1fs'%(time.time()-start_time))
f_write.write('\n')
# Store the result
fil = open(output_folder+'/result.pickle','wb')
pickle.dump(result_dic, fil)
fil.close()
f_write.close()
def main(args):
# Set up parameters.
alpha = 0.01
n_itr = 1500
# Set up the output folder.
output_folder = os.path.realpath(
'..') + '/results/result_univariate_' + args.output_folder
output_datafile = '/data3/martin/gtex_data/results_uni_covariate/result_' +\
args.output_folder + '.pickle'
if not os.path.exists(output_folder):
os.makedirs(output_folder)
else:
filelist = [os.remove(os.path.join(output_folder, f))\
for f in os.listdir(output_folder)]
# Load the data.
p, x, n_full, cate_name, cis_name = dl.load_GTEx(args.data_name)
# Logger.
logging.basicConfig(level=logging.INFO,format='%(message)s',\
filename=output_folder+'/result.log', filemode='w')
logger = logging.getLogger()
result_dic = {}
# An overview of the data
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(x[0:2, :]))
# Report the baseline methods.
n_rej, t_rej = md.bh_test(p, alpha=alpha, n_full=n_full, verbose=False)
logger.info('## BH, n_rej=%d, t_rej=%0.5f' % (n_rej, t_rej))
result_dic['bh'] = {'h_hat': p < t_rej}
n_rej, t_rej, pi0_hat = md.sbh_test(p,
alpha=alpha,
n_full=n_full,
verbose=False)
result_dic['sbh'] = {'h_hat': p < t_rej}
logger.info('## SBH, n_rej=%d, t_rej=%0.5f, pi0_hat=%0.3f\n' %
(n_rej, t_rej, pi0_hat))
# Analysis
md.adafdr_explore(p, x, alpha=alpha, n_full=n_full, vis_dim=None, cate_name=cate_name,\
output_folder=output_folder, h=None)
# Four covaraites seperately
cov_list = ['exp', 'maf', 'dist', 'chromotin', 'all']
for i_cov in range(5):
logger.info('Covariate: %s' % cov_list[i_cov])
if i_cov < 4:
temp_x = x[:, i_cov].reshape([-1, 1])
else:
temp_x = x
# Fast mode.
# output_folder_fast = output_folder + '_fast'
# if not os.path.exists(output_folder_fast):
# os.makedirs(output_folder_fast)
# else:
# filelist = [os.remove(os.path.join(output_folder_fast, f))\
# for f in os.listdir(output_folder_fast)]
output_folder_fast = None
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(temp_x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p,
temp_x,
K=5,
alpha=alpha,
h=None,
n_full=n_full,
n_itr=n_itr,
verbose=True,
output_folder=output_folder_fast,
random_state=0,
fast_mode=True)
n_rej = res['n_rej']
t_rej = res['threshold']
result_dic['nfdr (fast)_%d' % i_cov] = {'h_hat': p < t_rej}
logger.info(
'## AdaFDR (fast mode), feature=%d, n_rej1=%d, n_rej2=%d, n_rej_total=%d'
% (i_cov, n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time (fast mode): %0.1fs' %
(time.time() - start_time))
# Full mode.
logger.info('# p: %s' % str(p[0:2]))
logger.info('# x: %s' % str(temp_x[0:2, :]))
start_time = time.time()
res = md.adafdr_test(p, temp_x, K=5, alpha=alpha, h=None, n_full=n_full, n_itr=n_itr,\
verbose=True, output_folder=None, random_state=0,\
fast_mode=False, single_core=False)
n_rej = res['n_rej']
t_rej = res['threshold']
result_dic['nfdr_%d' % i_cov] = {'h_hat': p < t_rej}
logger.info('## nfdr2, n_rej1=%d, n_rej2=%d, n_rej_total=%d' %
(n_rej[0], n_rej[1], n_rej[0] + n_rej[1]))
logger.info('## Total time: %0.1fs' % (time.time() - start_time))
logger.info(' ')
# Store the result
fil = open(output_datafile, 'wb')
pickle.dump(result_dic, fil)
fil.close()