def correlateEigenvectorWithGeneTrack(pMatrix, pEigenvector, pGeneTrack):
'''
This function correlates the eigenvectors per chromosome with the gene
density. If the correlation is negative, the eigenvector values are
multiplied with -1.
'''
file_h = opener(pGeneTrack)
bed = ReadBed(file_h)
gene_occurrence = np.zeros(len(pMatrix.cut_intervals))
gene_occurrence_per_chr = {}
chromosome_list = pMatrix.getChrNames()
for interval in bed:
chromosome_name = interval.chromosome
if chromosome_name not in chromosome_list:
continue
# in which bin of the Hi-C matrix is the given gene?
bin_id = pMatrix.getRegionBinRange(interval.chromosome, interval.start,
interval.end)
# add +1 for one gene occurrence in this bin
gene_occurrence[bin_id[1]] += 1
for chromosome in chromosome_list:
# where is the start and the end bin of a chromosome?
bin_id = pMatrix.getChrBinRange(chromosome)
gene_occurrence_per_chr[chromosome] = \
gene_occurrence[bin_id[0]:bin_id[1]]
# change from [[1,2], [3,4], [5,6]] to [[1,3,5],[2,4,6]]
pEigenvector = np.array(pEigenvector).real.transpose()
# correlate gene density and eigenvector values.
# if positive correlation, do nothing, if negative, flip the values.
# computed per chromosome
for chromosome in chromosome_list:
bin_id = pMatrix.getChrBinRange(chromosome)
for i, eigenvector in enumerate(pEigenvector):
_correlation = pearsonr(eigenvector[bin_id[0]:bin_id[1]].real,
gene_occurrence_per_chr[chromosome])
if _correlation[0] < 0:
eigenvector[bin_id[0]:bin_id[1]] = np.negative(
eigenvector[bin_id[0]:bin_id[1]])
return np.array(pEigenvector).transpose()
def correlateEigenvectorWithGeneTrack(pMatrix, pEigenvector, pGeneTrack):
'''
This function correlates the eigenvectors per chromosome with the gene
density. If the correlation is negative, the eigenvector values are
multiplied with -1.
'''
log.debug('correlate eigenvector!')
file_h = opener(pGeneTrack)
bed = ReadBed(file_h)
gene_occurrence = np.zeros(len(pMatrix.cut_intervals))
gene_occurrence_per_chr = {}
chromosome_list = pMatrix.getChrNames()
for interval in bed:
chromosome_name = interval.chromosome
if chromosome_name not in chromosome_list:
continue
if interval.start > pMatrix.get_chromosome_sizes()[chromosome_name]:
log.warning(
'Your chromosome sizes do not match the chromosome sizes of the extraTrack data!'
)
log.warning(
'Your chromosome {}; Size {}. ExtraTrack data {} {} {}'.format(
chromosome_name,
pMatrix.get_chromosome_sizes()[chromosome_name],
interval.chromosome, interval.start, interval.end))
log.warning(
'Please create your interaction matrix with a chromosome size file! However, if the sizes are intended and it is accepted that certain regions are not part of the correlation, you can ignore this message.'
)
continue
# in which bin of the Hi-C matrix is the given gene?
bin_id = pMatrix.getRegionBinRange(interval.chromosome, interval.start,
interval.end)
if bin_id is None:
log.warning(
'Your chromosome sizes do not match the chromosome sizes of the extraTrack data!'
)
log.warning(
'Your chromosome {}; Size {}. ExtraTrack data {} {} {}'.format(
chromosome_name,
pMatrix.get_chromosome_sizes()[chromosome_name],
interval.chromosome, interval.start, interval.end))
log.warning(
'Please create your interaction matrix with a chromosome size file! However, if the sizes are intended and it is accepted that certain regions are not part of the correlation, you can ignore this message.'
)
continue
# add +1 for one gene occurrence in this bin
gene_occurrence[bin_id[1]] += 1
for chromosome in chromosome_list:
# where is the start and the end bin of a chromosome?
bin_id = pMatrix.getChrBinRange(chromosome)
gene_occurrence_per_chr[chromosome] = \
gene_occurrence[bin_id[0]:bin_id[1]]
# change from [[1,2], [3,4], [5,6]] to [[1,3,5],[2,4,6]]
pEigenvector = np.array(pEigenvector).real.transpose()
# correlate gene density and eigenvector values.
# if positive correlation, do nothing, if negative, flip the values.
# computed per chromosome
for chromosome in chromosome_list:
bin_id = pMatrix.getChrBinRange(chromosome)
for i, eigenvector in enumerate(pEigenvector):
_correlation = pearsonr(
np.array(eigenvector[bin_id[0]:bin_id[1]]).real,
gene_occurrence_per_chr[chromosome])
if _correlation[0] < 0:
eigenvector[bin_id[0]:bin_id[1]] = np.negative(
eigenvector[bin_id[0]:bin_id[1]])
# log.debug('correlated to {}!'.format(_correlation[0]))
return np.array(pEigenvector).transpose()