def make_fig(filename, hemi, surf, atlas_name, colour_func, output_dir, missing=(0.5, 0.5, 0.5)):
"""
Make a brain figure.
filename : path to the csv that contains the brain region names and the
associated values.
hemi : 'lh' or 'rh'.
surf : 'inflated' or 'pial'
atlas_name : 'Destrieux' or 'DSK'.
colour_func : function object that creates a colour dictionary based on the
data inside the csv.
output_dir : name of output directory (put inside 'outputs').
"""
if colour_func is not None:
min, max, cd = colour_func(filename, atlas_name)
else:
cd = {}
regions = atlas.get_atlas_labels(atlas_name)
for region in regions:
if region in cd:
c = cd[region]
load.load_surface(hemi, surf, atlas_name, region, c)
else:
if missing is not None:
c = missing
print "Warning: '%s' region missing (%s atlas)." % (region, atlas_name)
load.load_surface(hemi, surf, atlas_name, region, c)
# reposition the camera
angles = [[180, -90, -70], [180, 90, 70], [180, 0, 0], [0, 180, 0]]
fpath = paths.pj(paths.OUTPUTS_DIR, output_dir)
if not os.path.exists(fpath):
os.makedirs(fpath)
for a, e, r in angles:
f = paths.pj(fpath, "%i_%i_%i.png" % (a, e, r))
mlab.view(a, e, roll=r)
mlab.savefig(f, magnification=3)
def grab_phi2_matrix(ped_dir=paths.pj(paths.data_dir, 'newest_ped_files')):
"""
Returns a square matrix containing phi2 coefficients for all subjects in
the sample based on the pedigree files in the given directory, with the
singletons removed. Defaults to the oldest pedigree file that is used for
linkage.
"""
# load the phi2 sparse matrix
phi2_path = paths.pj(ped_dir, 'phi2.gz')
df = pd.read_table(
gzip.open(phi2_path, 'rb'), sep=' ', skipinitialspace=True
)
df.columns = ['id0', 'id1', 'phi', 'd7']
# convert to square matrix
n = df.id0.nunique()
R = np.zeros((n+1, n+1))
for ix, row in df.iterrows():
R[row.id0, row.id1] = row.phi
R[row.id1, row.id0] = row.phi
R = pd.DataFrame(R)
R.drop(0, axis=0, inplace=True)
R.drop(0, axis=1, inplace=True)
# label rows and columns according to subject ids
ped_path = paths.pj(ped_dir, 'pedigree.csv')
ids = pd.read_csv(ped_path, index_col=0).id
R.columns = ids
R = pd.concat([R, pd.DataFrame(ids)], axis=1)
R.set_index('id', inplace=True)
# remove singletons
_R = R.replace(1, np.nan)
rmax = _R.max(axis=1)
singletons = rmax[rmax == 0].index.tolist()
subjects = [s for s in R.index.tolist() if s not in singletons]
R = R.loc[subjects, subjects]
# return the complete phi2 matrix
return R
def grab_traits(dic):
"""
Returns a pandas DataFrame containing the desired traits. The input is a
dictionary with keys representing the csv file names and entries being
lists of trait names within that csv.
"""
dfs = []
age_found = False
for csv, traits in dic.iteritems():
df = pandas.read_csv(paths.pj(paths.DATA_DIR, csv), index_col=0)
if age_found is False:
if 'age' in df.columns:
traits += ['age']
age_found = True
elif 'Age' in df.columns:
traits += ['Age']
age_found = True
dfs.append(df[traits])
if age_found is False:
warnings.warn("No 'age' trait found in the supplied CSVs.")
return pandas.concat(dfs, axis=1)
def make_phi2_vectors(pheno_file, out_file, remove_affected=True, ped_dir=None,
traits=None):
"""
Loads a phenotype file, and creates phi2 vectors for each of the
dichotomous traits within it. Optionally removes affected subjects.
"""
# load the phi2 matrix
if not ped_dir:
R = grab_phi2_matrix()
else:
R = grab_phi2_matrix(ped_dir)
# load the phenotype(s)
df = pd.read_csv(pheno_file, index_col=0)
if traits is None:
traits = [c for c in df.columns if df[c].nunique() == 2]
else:
traits = traits.split(',')
subjects = R.index.tolist()
df = df.loc[subjects, traits]
# make the phi2 vector(s)
vectors, names = [], []
for trait_name in traits:
trait = df[trait_name].dropna()
cases = trait[trait == 1].index.tolist()
_R = R.loc[subjects, cases]
vector = _R.max(axis=1)
if remove_affected is True or remove_affected != 'False':
vector[vector == 1] = np.nan
vectors.append(vector)
name = 'phi2_%s_%s' % (trait_name, ('incl', 'excl')[remove_affected])
names.append(name)
df2 = pd.concat(vectors, axis=1)
df2.columns = names
# save the phi2 vectors
df2.to_csv(paths.pj(paths.merv_dir, out_file), index=True)
"""
Grab the sats.
"""
__author__ = 'smathias'
import os
import pandas
import solarpy.py2solar.parse as parse
import solarpy.utils.paths as paths
from make_asym_jobs import structures
path = paths.pj(paths.OUTPUTS_DIR, 'asymmetries_mpi')
data = []
for structure in structures:
dic = {'name': structure}
for x in ['Bilateral_', 'Left_', 'Right_', 'Asym_', 'Abs_Asym_']:
_f = x + structure + '.sub'
f = paths.pj(path, [f for f in os.listdir(path) if _f == f[:len(_f)]][0])
_dic = {x + k: v for k, v in parse.uni_polyg(f).iteritems()}
dic.update(_dic)
_f = 'rhog_' + structure + '.sub'
f = paths.pj(path, [f for f in os.listdir(path) if _f in f][0])
_dic = parse.biv_polyg(f)
dic.update(_dic)
data.append(dic)
traits_outrm['age'] = traits['Age']
traits_outrm['icv'] = traits['etiv']
traits1 = traits_outrm
# Make jobs
univariates = traits1.columns.tolist()
univariates.remove('age')
univariates.remove('icv')
for trait in univariates:
make_jobs.make_single_job(
'asymmetries_mpi', trait, [trait], cov='age sex n_icv', old_peds=False
)
for chrom in xrange(1, 23):
make_jobs.make_single_job(
'asymmetries_yale_1', '%s_%i' % (trait, chrom), [trait], chrom=chrom
)
for structure in structures:
make_jobs.make_single_job(
'asymmetries_mpi', 'rhog_' + structure,
['Left_%s' % structure, 'Right_%s' % structure],
cov='age sex n_icv', tests='-testrhog', old_peds=False
)
for chrom in xrange(1, 23):
make_jobs.make_single_job(
'asymmetries_yale_1', 'rhog_%s_%i' %(structure, chrom),
['Left_%s' % struct, 'Right_%s' % struct], tests='-testrhog',
chrom=chrom
)
traits1.to_csv(paths.pj(paths.JOBS_DIR, 'asymmetries_mpi', '_traits.csv'), index_label='id')
traits.to_csv(paths.pj(paths.JOBS_DIR, 'asymmetries_yale', '_traits.csv'), index_label='id')
"""
Grab the sats.
"""
__author__ = 'smathias'
import os
import pandas
import solarpy.py2solar.parse as parse
import solarpy.utils.paths as paths
outputs_dir = paths.pj(paths.OUTPUTS_DIR, 'braincog')
output_files = os.listdir(outputs_dir)
entries = {}
print len(output_files)
for f in output_files:
dic = parse.uni_polyg(paths.pj(outputs_dir, f))
entries[f.split('.')[0]] = dic
df = pandas.DataFrame(entries).T
df = df.drop(['f', 'trait', 'n', 'se'], axis=1)
df.to_csv('h2rs.csv')
def make_single_job(jobs_name, job_name, traits, old_peds=True,
inorm=True, resid=False, **kwargs):
"""
Creates a submission file and tcl file for a given job, and places them in
the specific jobs directory. By default the job will calculate h2rs of the
traits. Other tasks, such as running linkage, are done by passing kwargs.
"""
# make directory for job family
job_path = paths.pj(paths.JOBS_DIR, jobs_name)
if not os.path.exists(job_path):
os.makedirs(job_path)
# make a sub file
ep = templates.EXPORT_PATH + job_name + '/'
jp = templates.JOBS_PATH + jobs_name
if old_peds is True:
pp = templates.OLD_PED_PATH
else:
pp = templates.NEW_PED_PATH
jn = job_name
rp = templates.RESULTS_PATH
dic = locals()
dic = {k: v for k, v in dic.iteritems() if len(k) == 2}
sub = templates.SUB.format(**dic)
open(paths.pj(job_path, job_name + '.sub'), 'w').write(sub)
# make a tcl file
if inorm is True:
s = 'define n_%s = inorm_%s'
inorm_st = ''.join(s % (trait, trait) for trait in traits)
traits_st = ' '.join('n_%s' % trait for trait in traits)
else:
inorm_st = ''
traits_st = ' '.join(trait for trait in traits)
if 'cov' in kwargs:
if kwargs['cov'] is not None:
cov_st = 'cov ' + kwargs['cov']
else:
cov_st = ''
else:
cov_st = templates.COV
if resid is True:
resid_st = templates.RESID % job_name
else:
resid_st = ''
if 'tests' in kwargs:
tests_st = kwargs['tests']
else:
tests_st = ''
if 'chrom' in kwargs:
chrom = kwargs['chrom']
mibd = paths.MIBD_DIR_TX
if 'mibd' in kwargs:
mibd = kwargs['mibd']
if mibd == 'local':
mibd = paths.MIBD_DIR
multipoint_st = templates.MULTIPOINT % (chrom)
else:
multipoint_st = ''
job_name_st = job_name
dic = locals()
dic = {k[:-3]: v for k, v in dic.items() if '_st' in k}
print dic
tcl = templates.TCL_POLYG.substitute(**dic)
open(paths.pj(job_path, job_name + '.tcl'), 'w').write(tcl)
__author__ = 'smathias'
from itertools import product
import numpy as np
import pandas
import solarpy.utils.paths as paths
import solarpy.utils.qc as qc
import solarpy.py2solar.make_jobs as make_jobs
import matplotlib.pyplot as plt
import seaborn as sns
p = lambda f: paths.pj(paths.DATA_DIR, f)
fs = [
'all_facs_cog.csv', 'etiv4sam.csv', 'freesurfer.csv', 'gobs.csv',
'stan.csv'
]
dfs = [pandas.read_csv(p(f), index_col=0) for f in fs]
df = pandas.concat(dfs, axis=1)
wanted_cols = [
'GCOG_VM',
'GCOG_WM',
'GCOG_SM',
'GCOG_EF',
'GCOG_G',
'BIFG_VM',
'BIFG_WM',
'BIFG_SM',
"""
Grab the residualised traits.
"""
__author__ = 'smathias'
import os
import pandas
import solarpy.utils.paths as paths
results_dir = paths.pj(paths.RESULTS_DIR, 'braincog')
# results_dir = paths.RESULTS_DIR
folders = os.listdir(results_dir)
data = []
for folder in folders:
print folder
filename = paths.pj(results_dir, folder, 'resid_%s.csv' % folder)
s = pandas.read_csv(filename, index_col=0)['residual']
s.name = folder
data.append(s)
df = pandas.concat(data, axis=1)
df.to_csv('_traits_r.csv', index_label='id')