def _obs_exp(pSubmatrix):
obs_exp_matrix_ = obs_exp_matrix(pSubmatrix)
obs_exp_matrix_ = convertNansToZeros(csr_matrix(obs_exp_matrix_))
obs_exp_matrix_ = convertInfsToZeros(csr_matrix(obs_exp_matrix_))
# if len(obs_exp_matrix_.data) == 0:
# return np.array([[]])
return obs_exp_matrix_ # .todense()
def _obs_exp(pSubmatrix):
obs_exp_matrix_ = obs_exp_matrix(pSubmatrix)
obs_exp_matrix_ = convertNansToZeros(csr_matrix(obs_exp_matrix_))
obs_exp_matrix_ = convertInfsToZeros(csr_matrix(obs_exp_matrix_))
# log.error('obs_exp_matrix_.data {}'.format(obs_exp_matrix_.data))
# if len(obs_exp_matrix_.data) == 0:
# log.debug('No data!')
# return np.array([[]])
return obs_exp_matrix_ # .todense()
def compute_loops(pHiCMatrix, pRegion, pArgs, pIsCooler, pQueue=None):
"""
Master function to compute the loops for one chromosome.
- Removes all regions greater maxLoopSize
- Calls
- Writes computed loops to a bedgraph file
Input:
- pHiCMatrix: Hi-C interaction matrix object
- pRegion: Chromosome name
- pArgs: Argparser object
- pIsCooler: True / False if matrix is stored in a .cool file
- pQueue: Queue object for multiprocessing communication with parent process
"""
try:
if pQueue is not None:
if pIsCooler:
pHiCMatrix = hm.hiCMatrix(pMatrixFile=pArgs.matrix, pChrnameList=[pRegion], pDistance=pArgs.maxLoopDistance, pNoIntervalTree=True, pUpperTriangleOnly=False)
else:
pHiCMatrix = hm.hiCMatrix(pMatrixFile=pArgs.matrix, pChrnameList=[pRegion], pDistance=pArgs.maxLoopDistance, pNoIntervalTree=False, pUpperTriangleOnly=False)
if not pIsCooler:
# cooler files load only what is necessary.
pHiCMatrix.keepOnlyTheseChr([pRegion])
max_loop_distance = pArgs.maxLoopDistance / pHiCMatrix.getBinSize()
instances, features = pHiCMatrix.matrix.nonzero()
distances = np.absolute(instances - features)
mask = distances > max_loop_distance
pHiCMatrix.matrix.data[mask] = 0
pHiCMatrix.matrix.eliminate_zeros()
if len(pHiCMatrix.matrix.data) == 0:
pQueue.put([None])
return
if pHiCMatrix.matrix.shape[0] < 5 or pHiCMatrix.matrix.shape[1] < 5:
log.debug('Computed loops for {}: 0'.format(pRegion))
if pQueue is None:
return None
else:
pQueue.put([None])
return
if pArgs.windowSize is None:
bin_size = pHiCMatrix.getBinSize()
if 0 < bin_size <= 5000:
pArgs.windowSize = 10
elif 5000 < bin_size <= 10000:
pArgs.windowSize = 5
elif 10000 < bin_size <= 25000:
pArgs.windowSize = 5
elif 25000 < bin_size <= 50000:
pArgs.windowSize = 5
else:
pArgs.windowSize = 5
log.debug('Setting window size to: {}'.format(pArgs.windowSize))
if pArgs.peakWidth is None:
pArgs.peakWidth = pArgs.windowSize - 3
log.debug('Setting peak width to: {}'.format(pArgs.peakWidth))
pHiCMatrix.matrix = triu(pHiCMatrix.matrix, format='csr')
pHiCMatrix.matrix.eliminate_zeros()
# log.debug('candidates region {} {}'.format(
# pRegion, len(pHiCMatrix.matrix.data)))
# delete main diagonal
instances, features = pHiCMatrix.matrix.nonzero()
distances = np.absolute(instances - features)
mask = distances == 0
pHiCMatrix.matrix.data[mask] = 0
pHiCMatrix.matrix.eliminate_zeros()
del instances
del features
del mask
del distances
if pArgs.expected == 'mean':
obs_exp_csr_matrix = obs_exp_matrix(pHiCMatrix.matrix, pInplace=False, pToEpsilon=True, pThreads=pArgs.threadsPerChromosome)
elif pArgs.expected == 'mean_nonzero':
obs_exp_csr_matrix = obs_exp_matrix_non_zero(pHiCMatrix.matrix, ligation_factor=False, pInplace=False, pToEpsilon=True, pThreads=pArgs.threadsPerChromosome)
elif pArgs.expected == 'mean_nonzero_ligation':
obs_exp_csr_matrix = obs_exp_matrix_non_zero(pHiCMatrix.matrix, ligation_factor=True, pInplace=False, pToEpsilon=True, pThreads=pArgs.threadsPerChromosome)
if not isinstance(obs_exp_csr_matrix, csr_matrix):
if pQueue is None:
return None
else:
pQueue.put([None])
return
pHiCMatrix.matrix.eliminate_zeros()
obs_exp_csr_matrix.eliminate_zeros()
if len(pHiCMatrix.matrix.data) != len(obs_exp_csr_matrix.data):
if pQueue is None:
return None
else:
pQueue.put([None])
return
# handle pValuePreselection
try:
pArgs.pValuePreselection = float(pArgs.pValuePreselection)
except Exception:
pArgs.pValuePreselection = read_threshold_file(pArgs.pValuePreselection)
candidates, pValueList = compute_long_range_contacts(pHiCMatrix,
obs_exp_csr_matrix,
pArgs.windowSize,
pArgs.pValue,
pArgs.peakWidth,
pArgs.pValuePreselection,
pArgs.peakInteractionsThreshold,
pArgs.obsExpThreshold,
pArgs.threadsPerChromosome)
if candidates is None:
log.info('Computed loops for {}: 0'.format(pRegion))
if pQueue is None:
return None
else:
pQueue.put([None])
return
elif 'Fail: ' in candidates and pQueue is not None:
pQueue.put(candidates)
return
elif 'Fail: ' in candidates and pQueue is None:
return candidates
mapped_loops = cluster_to_genome_position_mapping(
pHiCMatrix, candidates, pValueList, pArgs.maxLoopDistance)
del pHiCMatrix
del candidates
log.debug('Computed loops for {}: {}'.format(pRegion, len(mapped_loops)))
except Exception as exp:
if pQueue is not None:
pQueue.put('Fail: ' + str(exp) + traceback.format_exc())
return
else:
return 'Fail: ' + str(exp) + traceback.format_exc()
if pQueue is None:
return mapped_loops
else:
pQueue.put([mapped_loops])
return
def _obs_exp(pSubmatrix):
obs_exp_matrix_ = obs_exp_matrix(pSubmatrix)
obs_exp_matrix_ = convertNansToZeros(csr_matrix(obs_exp_matrix_))
obs_exp_matrix_ = convertInfsToZeros(csr_matrix(obs_exp_matrix_)).todense()
return obs_exp_matrix_