def load_data(ch, dataset, snpsubset, snpruns, testpat, trainpat):
"""
Loading data from files into X and y matrices.
Selection of patients not being in test set and SNPs present in subset-file.
"""
snplist = {name: [] for name in dataset.keys()}
if snpsubset is not None:
for name in dataset.keys():
cc = open('%s%s/%s_snps_chr%d_%d.txt' % (dataset[name], snpsubset, snpsubset, ch, snpruns[name]), 'r')
for line in cc:
snplist[name].append(int(line.split()[0]))
cc.close()
snp = len(snplist[name])
else:
if len(dataset) > 1:
raise exceptions.NoParameterError('subset', 'There is more than one given data set, but subset of SNPs ' +
'is not given.')
else:
cc = open('%smatrices/genome_stats.txt' % list(dataset.values())[0], 'r')
snp = None
for line in cc:
if line.startswith('%d\t' % ch):
snp = int(line.split()[1])
break
cc.close()
if 'snp' is None:
raise exceptions.OtherError('There is no information about chromosome %d in %sgenome_stats.txt file'
% (ch, list(dataset.values())[0]))
snplist[next(iter(dataset.keys()))] = list(range(snp))
return read_Xs(ch, dataset, snp, snplist, testpat, trainpat)
def check_borutarun(directory, run, perc):
with open('%sboruta/boruta_runs.txt' % directory, 'r') as file:
for line in file:
if line.startswith(str(run) + '\t'):
line = line.split()
snpsubset = line[5]
if snpsubset == 'None':
snpsubset = None
snprun = None
else:
snprun = int(line[6].split('+')[0])
perc_list = line[8].split(',')
if len(perc_list) > 1 and perc is None:
raise exceptions.NoParameterError(
'perc',
'There is more than one perc value for given boruta run.'
)
elif perc is None:
perc = int(perc[0])
break
if 'snpsubset' not in locals():
raise exceptions.WrongValueError(
'run', run, 'Run number %d was not conducted' % run)
return perc, snpsubset, snprun
'%d\t%s\t%s\t%s\t%d\n' % (runs[setname], setname, ', '.join([
k for k in dataset.keys() if k != setname
]), funcs.make_chrstr(chrlist), shared_snps))
return shared_snps
dataset = OrderedDict()
chrlist = [i for i in range(1, 24)]
fixed = False
run = None
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
if sys.argv[q] == '-chr':
chrlist = funcs.read_chrstr(sys.argv[q + 1])
if sys.argv[q] == '-run':
run = int(sys.argv[q + 1])
if sys.argv[q] == '-fixed':
fixed = True
found = find_shared(dataset, chrlist, fixed, run)
print('%d shared SNPs found!' % found)
dir = './'
for q in range(len(sys.argv)):
if sys.argv[q] == '-dir':
dir = sys.argv[q + 1]
if sys.argv[q] == '-indir':
indir = sys.argv[q + 1]
if sys.argv[q] == '-outdir':
outdir = sys.argv[q + 1]
if sys.argv[q] == '-dataset':
dataset = sys.argv[q + 1]
if sys.argv[q] == '-diagdir':
diagdir = sys.argv[q + 1]
if 'dataset' not in globals():
raise exceptions.NoParameterError('dataset', 'e.g. adni or rosmap')
if 'diagdir' not in globals():
diagdir = '%sfiles/' % dir
if 'indir' not in globals():
indir = '%smatrices/' % dir
if 'outdir' not in globals():
outdir = '%smatrices/' % dir
if dataset == 'test':
dfiles = ['test_diagnoses.tsv']
dd = test_mapping(dfiles, diagdir)
if dataset == 'adni':
dfiles = ['dxsum.csv']
dataset = OrderedDict()
chrlist = [i for i in range(1, 24)]
fixed = False
run = None
borutaruns = None
perc = 90
thresh = 0.1
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
if sys.argv[q] == '-chr':
chrlist = funcs.read_chrstr(sys.argv[q + 1])
if sys.argv[q] == '-run':
run = int(sys.argv[q + 1])
if sys.argv[q] == '-fixed':
fixed = True
if sys.argv[q] == '-perc':
perc = int(sys.argv[q + 1])
if sys.argv[q] == '-borutarun':
if sys.argv[q + 1] in dataset.keys():
if borutaruns is None:
borutaruns = OrderedDict()
try:
borutaruns[sys.argv[q + 1]] = int(sys.argv[q + 2])
chs.sort()
for ch in chs:
stats.write('%d\t%d\t%d\n' % (ch, snps[str(ch)], pat))
stats.close()
indir = './'
for q in range(len(sys.argv)):
if sys.argv[q] == '-dbsnp':
dbsnp = sys.argv[q + 1]
if sys.argv[q] == '-plink':
plink = sys.argv[q + 1]
if sys.argv[q] == '-indir':
indir = sys.argv[q + 1]
if sys.argv[q] == '-outdir':
outdir = sys.argv[q + 1]
if 'outdir' not in globals():
outdir = indir
if 'plink' not in globals():
raise exceptions.NoParameterError('plink', 'name of plink files')
if 'dbsnp' not in globals():
raise exceptions.NoParameterError(
'dbsnp',
'name of file with list of SNPs assigned to their reference values')
pat = make_pid(plink, indir, outdir)
snps = make_ref(dbsnp, plink, indir, outdir)
genome_stats(pat, snps, outdir)
continuation = False
makey = False
cv = None
newforest = False
frombed = False
num_cores = None
method = 'rforest'
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
continue
if sys.argv[q] == '-testset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
testset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of test set should appear a directory to folder with it.'
)
continue
if sys.argv[q] == '-test':
testsize = float(sys.argv[q + 1])
dataset = OrderedDict()
procs = 1
chrlist = [i for i in range(1, 24)]
snpsubset = None
snpsubset = None
snpruns = None
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
continue
if sys.argv[q] == '-outdir':
if sys.argv[q + 1][0] in ['.', '~', '/']:
outdir = sys.argv[q + 1]
else:
raise exceptions.NoParameterError(
'outdir',
'After -outdir should appear a directory to output folder.')
continue
if sys.argv[q] == '-procs':
procs = int(sys.argv[q + 1])
continue
dataset = []
control_str = ['healthy', 'control', 'normal', 'NL', 'CN', '0']
case_str = ['ill', 'case', 'AD', 'alzheimer', '1']
all_str = ['all', 'whole', 'every']
kdeplot = False
clustermap = False
dendrogram = False
for q in range(len(sys.argv)):
if sys.argv[q].startswith('-dataset'):
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset.append([sys.argv[q + 1], sys.argv[q + 2]])
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
if sys.argv[q] == '-matrix':
if sys.argv[q + 1][0] in ['.', '~', '/']:
sims = np.load(sys.argv[q + 1])
else:
raise exceptions.NoParameterError(
'directory',
'After -matrix should appear a directory to similarity' +
'matrix written in .npy file.')
if sys.argv[q].startswith('-seta') or sys.argv[q].startswith('-setA'):
seta = add_to_set(sys.argv, seta, q)
except ValueError:
v = -1
X[j, i + 1] = v
np.savetxt('%sX_chr%s.csv' % (outdir, ch),
X,
fmt='%d',
delimiter=',',
header=',' + ','.join(list(map(str, range(snp)))),
comments='')
o.close()
s.close()
return "%s\t%d\t%d" % (ch, snp, pat)
ch = '1'
outdir = './'
for q in range(len(sys.argv)):
if sys.argv[q] == '-chr':
ch = sys.argv[q + 1]
if sys.argv[q] == '-input':
inp = sys.argv[q + 1]
if sys.argv[q] == '-outdir':
outdir = sys.argv[q + 1]
if 'inp' not in globals():
raise exceptions.NoParameterError('inp', 'name of input file')
print(vcf_to_matrix(ch, inp, outdir))
action='store',
type=str,
metavar='NAME',
default=None,
help='name of SNP subset')
parser.add_argument('-snprun', action=SnprunAction, nargs='*', metavar='')
args = parser.parse_args()
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
dataset[sys.argv[q + 1]] = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
continue
if sys.argv[q] == '-test':
testsize = float(sys.argv[q + 1])
testsize_given = True
continue
if sys.argv[q] == '-perc':
perc = list(map(int, sys.argv[q + 1].split(',')))
perc_given = True
continue
if sys.argv[q] == '-snpsubset':
snpsubset = sys.argv[q + 1]
s += 1
n = nn = False
v1 = v2 = -2
np.save('%smatrix_chr%s.npy' % (outdir, ch), matrix)
done += 2 * snp
stats.close()
pedfile.close()
indir = './'
overwrite = False
for q in range(len(sys.argv)):
if sys.argv[q] == '-plink':
plink = sys.argv[q + 1]
if sys.argv[q] == '-indir':
indir = sys.argv[q + 1]
if sys.argv[q] == '-outdir':
outdir = sys.argv[q + 1]
if sys.argv[q] == '-overwrite':
overwrite = True
if 'outdir' not in globals():
outdir = indir
if 'plink' not in globals():
raise exceptions.NoParameterError('plink', 'name of plink files')
snps_val = write_snps_list(plink, indir, outdir, overwrite)
write_matrix(plink, indir, outdir, snps_val)
print('Ill: %d' % diagnoses['1'])
return 0
run = None
fixed = False
for q in range(len(sys.argv)):
if sys.argv[q] == '-dataset':
if sys.argv[q + 2][0] in ['.', '~', '/']:
name = sys.argv[q + 1]
dir = sys.argv[q + 2]
else:
raise exceptions.NoParameterError(
'directory',
'After name of data set should appear a directory to folder with it.'
)
if sys.argv[q] == '-outdir':
if sys.argv[q + 1][0] in ['.', '~', '/']:
outdir = sys.argv[q + 1]
else:
raise exceptions.NoParameterError(
'outdir',
'After -outdir should appear a directory to output folder.')
if sys.argv[q] == '-lower':
lower = float(sys.argv[q + 1])
if sys.argv[q] == '-upper':
upper = float(sys.argv[q + 1])