def main():
X, xheader, yheader = mf.dekker_2_numpy_matrix(args.i)
# print X, xheader, yheader
print "Matrix shape: " + str(np.shape(X))
# Step 1
# Rows and columns for which more than 30% of the entries were either undefined or zeros
# were removed from the matrix
print "Removing NAs..."
X_step1, xheader, yheader = mf.remove_NA_zeros(X, xheader, yheader, 0.65)
# print X_step1, xheader, yheader
print "Matrix shape: " + str(np.shape(X_step1))
# Step 2
# Determine whether to transpose matrix
if args.c:
# Transpose matrix
print "Transposing matrix..."
X_step2 = np.transpose(X_step1)
else:
X_step2 = X_step1
# Step 3
# Transform matrix to zscores by row
print "Transforming rows to zscores..."
X_step3 = stats.zscore(X_step2, axis=1)
# Step 4
# Train hmm model
print "Training HMM"
model = hmm.GaussianHMM(n_components=5, covariance_type="diag", n_iter=1000)
model.fit(X_step3)
print "HMM output"
classes = model.predict(X_step3)
if args.c:
print "# of columns: " + str(len(xheader))
print "# of class entries: " + str(len(classes))
OUT = open("even_compartments.tab", "w")
for i in range(len(xheader)):
OUT.write(xheader[i] + "\t" + str(classes[i]) + "\n")
OUT.close()
else:
print "# of rows: " + str(len(yheader))
print "# of class entries: " + str(len(classes))
OUT = open("odd_compartments.tab", "w")
for i in range(len(yheader)):
OUT.write(yheader[i] + "\t" + str(classes[i]) + "\n")
OUT.close()
def main():
X, xheader, yheader = mf.dekker_2_numpy_matrix(args.i)
#print X, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X))
# Step 1
# Rows and columns for which more than 30% of the entries were either undefined or zeros
# were removed from the matrix
print 'Removing NAs...'
X_step1, xheader, yheader = mf.remove_NA_zeros(X, xheader, yheader,.65)
#print X_step1, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X_step1))
#Calculate contact enrichment
contact_enrich = np.zeros((5,5), dtype=object)
for i in range(len(contact_enrich)):
for j in range(len(contact_enrich[i])):
contact_enrich[i,j] = []
ODD_FH = open(args.o, 'r')
EVEN_FH = open(args.e, 'r')
odd = []
even = []
for line in ODD_FH:
odd.append(int(line.split()[1]))
ODD_FH.close()
for line in EVEN_FH:
even.append(int(line.split()[1]))
EVEN_FH.close()
# Size check
if len(odd) != len(X_step1) or len(even) != len(X_step1[0,:]):
print 'ERROR: size mismatches exist'
quit()
for i in range(len(X_step1)):
if i % 1000 == 0:
print 'On row: ' + str(i)
for j in range(len(X_step1[0,:])):
odd_class = odd[i]
even_class = even[j]
contact_enrich[odd_class, even_class].append(X_step1[i,j])
for i in range(len(contact_enrich)):
for j in range(len(contact_enrich[0,:])):
contact_enrich[i,j] = np.mean(contact_enrich[i,j])
np.savetxt('contact_enrichment.tab', contact_enrich, delimiter='\t', fmt='%1.2f')
def main():
parser = argparse.ArgumentParser(
description=
'Calculate contact enrichment to merge odd/even subcompartment calls')
parser.add_argument('-i',
help='input interaction matrix file',
type=str,
required=True)
parser.add_argument(
'-o',
help='odd chrom subcompartment calls (ex. odd_compartments.tab)',
type=str,
required=True)
parser.add_argument(
'-e',
help='even chrom subcompartment calls (ex. even_compartments.tab)',
type=str,
required=True)
args = parser.parse_args()
X, xheader, yheader = mf.dekker_2_numpy_matrix(args.i)
#print X, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X))
# Step 1
# Rows and columns for which more than 30% of the entries were either undefined or zeros
# were removed from the matrix
print 'Removing NAs...'
X_step1, xheader, yheader = mf.remove_NA_zeros(X, xheader, yheader, .65)
#print X_step1, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X_step1))
#Calculate contact enrichment
contact_enrich = np.zeros((5, 5), dtype=object)
for i in range(len(contact_enrich)):
for j in range(len(contact_enrich[i])):
contact_enrich[i, j] = []
ODD_FH = open(args.o, 'r')
EVEN_FH = open(args.e, 'r')
odd = []
even = []
for line in ODD_FH:
odd.append(int(line.split()[1]))
ODD_FH.close()
for line in EVEN_FH:
even.append(int(line.split()[1]))
EVEN_FH.close()
# Size check
if len(odd) != len(X_step1) or len(even) != len(X_step1[0, :]):
print 'ERROR: size mismatches exist'
quit()
for i in range(len(X_step1)):
if i % 1000 == 0:
print 'On row: ' + str(i)
for j in range(len(X_step1[0, :])):
odd_class = odd[i]
even_class = even[j]
contact_enrich[odd_class, even_class].append(X_step1[i, j])
for i in range(len(contact_enrich)):
for j in range(len(contact_enrich[0, :])):
contact_enrich[i, j] = np.mean(contact_enrich[i, j])
np.savetxt('contact_enrichment.tab',
contact_enrich,
delimiter='\t',
fmt='%1.2f')
def main():
parser = argparse.ArgumentParser(
description='Perform Rao et al. (2014) HMM subcompartment analysis')
parser.add_argument('-i',
help='input interaction matrix file',
type=str,
required=True)
parser.add_argument('-c',
help='run HMM on columns/even chromosomes)',
type=bool,
default=False)
args = parser.parse_args()
X, xheader, yheader = mf.dekker_2_numpy_matrix(args.i)
#print X, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X))
# Step 1
# Rows and columns for which more than 30% of the entries were either undefined or zeros
# were removed from the matrix
print 'Removing NAs...'
X_step1, xheader, yheader = mf.remove_NA_zeros(X, xheader, yheader, .65)
#print X_step1, xheader, yheader
print 'Matrix shape: ' + str(np.shape(X_step1))
# Step 2
# Determine whether to transpose matrix
if args.c:
# Transpose matrix
print 'Transposing matrix...'
X_step2 = np.transpose(X_step1)
else:
X_step2 = X_step1
# Step 3
# Transform matrix to zscores by row
print 'Transforming rows to zscores...'
X_step3 = stats.zscore(X_step2, axis=1)
# Step 4
# Train hmm model
print 'Training HMM'
model = hmm.GaussianHMM(n_components=5,
covariance_type='diag',
n_iter=1000)
model.fit(X_step3)
print 'HMM output'
classes = model.predict(X_step3)
if args.c:
print '# of columns: ' + str(len(xheader))
print '# of class entries: ' + str(len(classes))
OUT = open('even_compartments.tab', 'w')
for i in range(len(xheader)):
OUT.write(xheader[i] + '\t' + str(classes[i]) + '\n')
OUT.close()
else:
print '# of rows: ' + str(len(yheader))
print '# of class entries: ' + str(len(classes))
OUT = open('odd_compartments.tab', 'w')
for i in range(len(yheader)):
OUT.write(yheader[i] + '\t' + str(classes[i]) + '\n')
OUT.close()