def _compute_svmnormalvector((cache_dir, images, control_images,
normalization_name, preprocess_file, rfe)):
#try:
import numpy as np
import sys
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import RobustLinearNormalization, normalizations
from sklearn.svm import LinearSVC
from cpf.profiling.profile_svmnormalvector import _compute_rfe
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
normalizeddata, normalized_colnames, _ = cache.load(images, normalization=normalization)
control_data, control_colnames, _ = cache.load(control_images, normalization=normalization)
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
normalizeddata = preprocessor(normalizeddata)
control_data = preprocessor(control_data)
assert len(control_data) >= len(normalizeddata)
downsampled = control_data[np.random.randint(0, len(control_data), len(normalizeddata)), :]
x = np.vstack((normalizeddata, downsampled))
y = np.array([1] * len(normalizeddata) + [0] * len(downsampled))
clf = LinearSVC(C=1.0)
m = clf.fit(x, y)
normal_vector = m.coef_[0]
if rfe:
# Copy because it is immutable (normal_vector.flags.weriteable == False)
normal_vector = np.array(normal_vector)
normal_vector[~_compute_rfe(x, y)] = 0
return normal_vector
def _compute_ksstatistic((cache_dir, images, control_images,
normalization_name, preprocess_file)):
import numpy as np
import sys
import cpf
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import normalizations
from cpf.profiling.ks_2samp import ks_2samp
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
normalizeddata, variables, _ = cache.load(images,
normalization=normalization)
control_data, control_colnames, _ = cache.load(control_images,
normalization=normalization)
assert len(control_data) >= len(normalizeddata)
assert variables == control_colnames
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
normalizeddata = preprocessor(normalizeddata)
control_data = preprocessor(control_data)
variables = preprocessor.variables
#downsampled = control_data[np.random.randint(0, len(control_data), len(normalizeddata)), :]
m = len(variables)
profile = np.empty(m)
for j in range(m):
profile[j] = ks_2samp(control_data[:, j],
normalizeddata[:, j],
signed=True)[0]
return profile
def _compute_ksstatistic((cache_dir, images, control_images, normalization_name,
preprocess_file)):
import numpy as np
import sys
import cpf
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import normalizations
from cpf.profiling.ks_2samp import ks_2samp
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
normalizeddata, variables, _ = cache.load(images, normalization=normalization)
control_data, control_colnames, _ = cache.load(control_images, normalization=normalization)
assert len(control_data) >= len(normalizeddata)
assert variables == control_colnames
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
normalizeddata = preprocessor(normalizeddata)
control_data = preprocessor(control_data)
variables = preprocessor.variables
#downsampled = control_data[np.random.randint(0, len(control_data), len(normalizeddata)), :]
m = len(variables)
profile = np.empty(m)
for j in range(m):
profile[j] = ks_2samp(control_data[:, j], normalizeddata[:, j],
signed=True)[0]
return profile
def _compute_group_mean((cache_dir, images, normalization_name,
preprocess_file, method)):
try:
import numpy as np
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import normalizations
from scipy.stats import norm as Gaussian
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
data, colnames, _ = cache.load(images, normalization=normalization)
cellcount = np.ones(1) * data.shape[0]
if method == 'cellcount':
return cellcount
if len(data) == 0:
return np.empty(len(colnames)) * np.nan
data = data[~np.isnan(np.sum(data, 1)), :]
if len(data) == 0:
return np.empty(len(colnames)) * np.nan
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
data = preprocessor(data)
if method == 'mean':
return np.mean(data, axis=0)
elif method == 'mean+std':
return np.hstack((np.mean(data, axis=0), np.std(data, axis=0)))
elif method == 'mode':
return mode(data, axis=0)
elif method == 'median':
return np.median(data, axis=0)
elif method == 'median+mad':
c = Gaussian.ppf(3/4.)
d = np.median(data, axis=0)
return np.hstack((d,
np.median((np.fabs(data-d)) / c, axis=0)))
elif method == 'gmm2':
max_sample_size = 2000
if data.shape[0] > max_sample_size:
data = data[np.random.random_integers(0,data.shape[0]-1,size=max_sample_size),:]
from sklearn.decomposition import PCA
from sklearn.mixture import GMM
pca = PCA(n_components=0.99).fit(data)
pca_data = pca.transform(data)
#gmm = GMM(2, covariance_type='full', n_iter=100000, thresh=1e-7).fit(pca_data)
gmm = GMM(2, covariance_type='full').fit(pca_data)
return pca.inverse_transform(gmm.means_).flatten()
elif method == 'deciles':
return np.hstack(map(lambda d: np.percentile(data, d, axis=0), range(10,100,10)))
elif method == 'mean+deciles':
return np.hstack((np.mean(data, axis=0), np.hstack(map(lambda d: np.percentile(data, d, axis=0), range(10,100,10)))))
except: # catch *all* exceptions
from traceback import print_exc
import sys
print_exc(None, sys.stderr)
return None
def _compute_svmnormalvector((cache_dir, images, control_images,
normalization_name, preprocess_file, rfe)):
#try:
import numpy as np
import sys
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import RobustLinearNormalization, normalizations
from sklearn.svm import LinearSVC
from cpf.profiling.profile_svmnormalvector import _compute_rfe
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
normalizeddata, normalized_colnames, _ = cache.load(
images, normalization=normalization)
control_data, control_colnames, _ = cache.load(control_images,
normalization=normalization)
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
normalizeddata = preprocessor(normalizeddata)
control_data = preprocessor(control_data)
assert len(control_data) >= len(normalizeddata)
downsampled = control_data[
np.random.randint(0, len(control_data), len(normalizeddata)), :]
x = np.vstack((normalizeddata, downsampled))
y = np.array([1] * len(normalizeddata) + [0] * len(downsampled))
clf = LinearSVC(C=1.0)
m = clf.fit(x, y)
normal_vector = m.coef_[0]
if rfe:
# Copy because it is immutable (normal_vector.flags.weriteable == False)
normal_vector = np.array(normal_vector)
normal_vector[~_compute_rfe(x, y)] = 0
return normal_vector
def profile_mean(cache_dir,
group_name,
filter=None,
parallel=Uniprocessing(),
normalization=RobustLinearNormalization,
preprocess_file=None,
show_progress=True,
method='mean',
full_group_header=False):
group, colnames_group = cpf.db.group_map(group_name,
reverse=True,
filter=filter)
keys = group.keys()
parameters = [(cache_dir, group[g], normalization.__name__,
preprocess_file, method) for g in keys]
if "CPA_DEBUG" in os.environ:
DEBUG_NGROUPS = 5
logging.warning(
'In debug mode. Using only a few groups (n=%d) to create profile' %
DEBUG_NGROUPS)
parameters = parameters[0:DEBUG_NGROUPS]
keys = keys[0:DEBUG_NGROUPS]
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
variables = preprocessor.variables
else:
cache = Cache(cache_dir)
variables = normalization(cache).colnames
if method == 'mean+std':
variables = variables + ['std_' + v for v in variables]
elif method == 'median+mad':
variables = variables + ['mad_' + v for v in variables]
elif method == 'gmm2':
variables = ['m1_' + v
for v in variables] + ['m2_' + v for v in variables]
elif method == 'deciles':
variables = [
'decile_%02d_%s' % (dec, v) for dec in range(10, 100, 10)
for v in variables
]
elif method == 'mean+deciles':
variables = variables + [
'decile_%02d_%s' % (dec, v) for dec in range(10, 100, 10)
for v in variables
]
elif method == 'cellcount':
variables = ['Cells_Count']
return Profiles.compute(
keys,
variables,
_compute_group_mean,
parameters,
parallel=parallel,
group_name=group_name,
show_progress=show_progress,
group_header=colnames_group if full_group_header else None)
def profile_svmnormalvector(cache_dir,
group_name,
control_filter,
filter=None,
rfe=False,
job=None,
parallel=Uniprocessing(),
normalization=RobustLinearNormalization,
preprocess_file=None):
group, colnames_group = cpf.db.group_map(group_name,
reverse=True,
filter=filter)
control_images_by_plate = images_by_plate(control_filter)
plate_by_image = dict(
(row[:-2], row[-2]) for row in cpf.db.GetPlatesAndWellsPerImage())
def control_images(treated_images):
return [
r for image in treated_images
for r in control_images_by_plate[plate_by_image[image]]
]
keys = group.keys()
parameters = [(cache_dir, group[k], control_images(group[k]),
normalization.__name__, preprocess_file, rfe) for k in keys]
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
variables = preprocessor.variables
else:
cache = Cache(cache_dir)
variables = normalization(cache).colnames
if job:
i = job - 1
memoize(_compute_svmnormalvector)
else:
if memoization_dir is None:
fn = _compute_svmnormalvector
else:
fn = memoizer(_compute_svmnormalvector)
return Profiles.compute(keys,
variables,
fn,
parameters,
parallel=parallel,
group_name=group_name)
def profile_ksstatistic(cache_dir,
group_name,
control_filter,
plate_group,
filter=None,
parallel=Uniprocessing(),
normalization=RobustLinearNormalization,
preprocess_file=None):
group, colnames_group = cpf.db.group_map(group_name,
reverse=True,
filter=filter)
control_images_by_plate = images_by_plate(control_filter, plate_group)
plate_by_image = dict((row[:-2], tuple(row[-2:-1]))
for row in cpf.db.GetPlatesAndWellsPerImage())
def control_images(treated_images):
if plate_group is None:
return control_images_by_plate[None]
else:
return list(
set(r for image in treated_images
for r in control_images_by_plate[plate_by_image[image]]))
keys = group.keys()
parameters = [(cache_dir, group[k], control_images(group[k]),
normalization.__name__, preprocess_file) for k in keys]
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
variables = preprocessor.variables
else:
cache = Cache(cache_dir)
variables = normalization(cache).colnames
return Profiles.compute(keys,
variables,
_compute_ksstatistic,
parameters,
parallel=parallel,
group_name=group_name)
def _compute_group_mean((cache_dir, images, normalization_name,
preprocess_file, method)):
try:
import numpy as np
from cpf.profiling.cache import Cache
from cpf.profiling.normalization import normalizations
from scipy.stats import norm as Gaussian
cache = Cache(cache_dir)
normalization = normalizations[normalization_name]
data, colnames, _ = cache.load(images, normalization=normalization)
cellcount = np.ones(1) * data.shape[0]
if method == 'cellcount':
return cellcount
if len(data) == 0:
return np.empty(len(colnames)) * np.nan
data = data[~np.isnan(np.sum(data, 1)), :]
if len(data) == 0:
return np.empty(len(colnames)) * np.nan
if preprocess_file:
preprocessor = cpf.util.unpickle1(preprocess_file)
data = preprocessor(data)
if method == 'mean':
return np.mean(data, axis=0)
elif method == 'mean+std':
return np.hstack((np.mean(data, axis=0), np.std(data, axis=0)))
elif method == 'mode':
return mode(data, axis=0)
elif method == 'median':
return np.median(data, axis=0)
elif method == 'median+mad':
c = Gaussian.ppf(3 / 4.)
d = np.median(data, axis=0)
return np.hstack((d, np.median((np.fabs(data - d)) / c, axis=0)))
elif method == 'gmm2':
max_sample_size = 2000
if data.shape[0] > max_sample_size:
data = data[np.random.random_integers(
0, data.shape[0] - 1, size=max_sample_size), :]
from sklearn.decomposition import PCA
from sklearn.mixture import GMM
pca = PCA(n_components=0.99).fit(data)
pca_data = pca.transform(data)
#gmm = GMM(2, covariance_type='full', n_iter=100000, thresh=1e-7).fit(pca_data)
gmm = GMM(2, covariance_type='full').fit(pca_data)
return pca.inverse_transform(gmm.means_).flatten()
elif method == 'deciles':
return np.hstack(
map(lambda d: np.percentile(data, d, axis=0),
range(10, 100, 10)))
elif method == 'mean+deciles':
return np.hstack((np.mean(data, axis=0),
np.hstack(
map(lambda d: np.percentile(data, d, axis=0),
range(10, 100, 10)))))
except: # catch *all* exceptions
from traceback import print_exc
import sys
print_exc(None, sys.stderr)
return None
normalizations = dict((c.__name__, c) for c in [
RobustLinearNormalization, RobustStdNormalization, StdNormalization,
DummyNormalization
])
if __name__ == '__main__':
logging.basicConfig(level=logging.INFO)
parser = OptionParser(
"usage: %prog [-r] [-m method] PROPERTIES-FILE CACHE-DIR PREDICATE")
parser.add_option('-m',
'--method',
dest='method',
action='store',
default='RobustStdNormalization',
help='method')
options, args = parser.parse_args()
if len(args) != 3:
parser.error('Incorrect number of arguments')
properties_file, cache_dir, predicate = args
cpf.properties.LoadFile(properties_file)
from cpf.profiling.cache import Cache
cache = Cache(cache_dir)
normalizer = normalizations[options.method](cache)
normalizer._create_cache(predicate)