def geneate_biases_ICE(chromosome, matrix, bins, output):
chro_name = str(chromosome)
mid_points = (bins[:, 1] + bins[:, 2]) / 2
mid_points = mid_points.astype(int)
X, bias = normalization.ICE_normalization(matrix, output_bias=True)
bias = bias.flatten().astype(float)
with open(os.path.join(output + '_bias.txt'), 'w+') as f:
for mp, bs in zip(mid_points, bias):
line = '{}\t{}\t{}\n'.format(chro_name, mp, bs)
f.write(line)
f.close()
with open(os.path.join(output + '_bias.txt'), 'rb') as f_in:
with gzip.open(os.path.join(output + '_bias.txt.gz'), 'wb') as f_out:
shutil.copyfileobj(f_in, f_out)
f_out.close()
f_in.close()
os.remove(os.path.join(output + '_bias.txt'))
def plot_demo(source_dir, chromosome, model_name, data_type, resolution, start, end, destination_dir):
cool_file = os.path.join(source_dir, 'sample_{}_chr{}.cool'.format(model_name, chromosome))
hic = cooler.Cooler(cool_file)
start = max(0, int(start))
if end > hic.chromsizes['chr{}'.format(chromosome)]:
length = end - start
end = hic.chromsizes['chr{}'.format(chromosome)]
start = end - length
region = ('chr{}'.format(chromosome), start, end)
hic_mat = hic.matrix(balance=True).fetch(region)
hic_mat = normalization.ICE_normalization(hic_mat)
# hic_bins = hic.bins().fetch(region)
# weight = hic_bins['weight']
# filter_idx = np.array(np.where(weight==1)).flatten()
fig, ax0 = plt.subplots()
cmap = plt.get_cmap('RdBu_r')
hic_mat = filter_diag_boundary(hic_mat, diag_k=1, boundary_k=200)
Z = np.log1p(hic_mat)
bounds = np.append(np.arange(0,7,0.06), np.arange(7,12,0.3))
norm = matplotlib.colors.BoundaryNorm(boundaries=bounds, ncolors=256)
im = ax0.imshow(Z, cmap=cmap, norm=norm) #, vmin=0, vmax=8
fig.colorbar(im, ax=ax0, ticks=np.arange(0,8))
legend = {'ours': 'EnHiC', 'deephic': 'Deephic', 'hicsr':'HiCSR', 'low':'LR', 'high':'HR'}
name = model_name.split('_')[0]
if name == 'high' and 'down' in data_type:
ax0.set_title('{}'.format(legend[name]), fontsize=24)
else:
ax0.set_title('{}, {}'.format(legend[name], data_type), fontsize=24)
ax0.set_xlim(-1, hic_mat.shape[0])
ax0.set_ylim(-1, hic_mat.shape[1])
fig.tight_layout()
output = destination_dir
os.makedirs(output, exist_ok=True)
plt.savefig(os.path.join(output, 'demo_{}.pdf'.format(legend[name])), format='pdf')
plt.savefig(os.path.join(output, 'demo_{}.jpg'.format(legend[name])), format='jpg')
def plot_significant_interactions(source_dir, chromosome, model_name,
resolution, low_dis, up_dis, start, end):
cool_file = os.path.join(source_dir,
'{}_chr{}.cool'.format(model_name, chromosome))
hic = cooler.Cooler(cool_file)
start = max(0, int(start))
if end > hic.chromsizes['chr{}'.format(chromosome)]:
length = end - start
end = hic.chromsizes['chr{}'.format(chromosome)]
start = end - length
region = ('chr{}'.format(chromosome), start, end)
hic_mat = hic.matrix(balance=True).fetch(region)
hic_mat = normalization.ICE_normalization(hic_mat)
# hic_bins = hic.bins().fetch(region)
# weight = hic_bins['weight']
# filter_idx = np.array(np.where(weight==1)).flatten()
prefix = '{}_chr{}_{}_{}'.format(model_name, chromosome, start, end)
model_path = os.path.join(
source_dir, 'output_{}_{}'.format(start, end), prefix,
'FitHiC.spline_pass1.res10000.significances.txt.gz')
if not os.path.isfile(model_path):
return
model_data = pd.read_csv(model_path,
compression='gzip',
header=0,
sep='\t')
if model_data.empty:
return
model_si = extract_si(model_data)
q_idx = np.array(np.where(model_si[:, 2] < 0.05)).flatten()
si_x = np.floor((model_si[q_idx, 0].flatten() - start) / resolution)
si_y = np.floor((model_si[q_idx, 1].flatten() - start) / resolution)
fig, ax0 = plt.subplots()
cmap = plt.get_cmap('coolwarm')
hic_mat = filter_diag_boundary(hic_mat, diag_k=1, boundary_k=200)
Z = np.log1p(hic_mat)
im = ax0.imshow(Z, cmap=cmap, vmin=0, vmax=8)
fig.colorbar(im, ax=ax0)
ax0.scatter(si_x.flatten(),
si_y.flatten(),
color='#00ff00',
s=6,
edgecolors='none') # "turquoise"
legend = {
'ours': 'EnHiC',
'deephic': 'Deephic',
'hicsr': 'HiCSR',
'low': 'LR',
'high': 'HR'
}
name = model_name.split('_')[0]
ax0.set_title('{} log1p Scale'.format(legend[name]))
ax0.set_xlim(-1, hic_mat.shape[0])
ax0.set_ylim(-1, hic_mat.shape[1])
fig.tight_layout()
output = os.path.join(source_dir, 'figure', '{}_{}'.format(start, end))
os.makedirs(output, exist_ok=True)
plt.savefig(os.path.join(
output, '{}_chr{}_{}_{}.pdf'.format(model_name, chromosome, start,
end)),
format='pdf')
plt.savefig(os.path.join(
output, '{}_chr{}_{}_{}.jpg'.format(model_name, chromosome, start,
end)),
format='jpg')
from matplotlib import colors
from iced import datasets
from iced import filter
from iced import normalization
"""
Normalizing a contact count matrix.
"""
# Loading a sample dataset
counts, lengths = datasets.load_sample_yeast()
# Filtering and normalizing contact count data
normed = filter.filter_low_counts(counts, lengths=lengths, percentage=0.04)
normed = normalization.ICE_normalization(normed)
# Plotting the results using matplotlib
chromosomes = ["I", "II", "III", "IV", "V", "VI"]
fig, axes = plt.subplots(ncols=2, figsize=(12, 4))
axes[0].imshow(counts, cmap="Blues", norm=colors.SymLogNorm(1),
origin="bottom",
extent=(0, len(counts), 0, len(counts)))
[axes[0].axhline(i, linewidth=1, color="#000000") for i in lengths.cumsum()]
[axes[0].axvline(i, linewidth=1, color="#000000") for i in lengths.cumsum()]
axes[0].set_title("Raw contact counts")
m = axes[1].imshow(normed, cmap="Blues", norm=colors.SymLogNorm(1),
from iced import filter, normalization
import numpy as np
filePtr = './GSM1173492_Th1_ensemble/50kb/'
for i in range(1, 11):
counts = np.genfromtxt(filePtr + 'chr' + str(i) + '_50kb.txt',
delimiter=' ')
counts = filter.filter_low_counts(counts, percentage=0.04)
normed = normalization.ICE_normalization(counts)
np.savetxt(filePtr + 'Iced_chr' + str(i) + '_' + '_50kb.txt',
normed,
delimiter=',')
#
# The normalization is done in three step:
#
# 1. Normalize the data using LOIC, to remove GC, mappability, and other
# biases
# 2. Estimate the block biases due to copy number.
# 3. Remove the block biases from the LOIC-normalized contact counts
from iced import datasets
from iced import normalization
import matplotlib.pyplot as plt
from matplotlib import colors
counts, lengths, cnv = datasets.load_sample_cancer()
loic_normed = normalization.ICE_normalization(counts, counts_profile=cnv)
block_biases = normalization.estimate_block_biases(counts, lengths, cnv)
caic_normed = loic_normed / block_biases
###############################################################################
# Visualizing the results using Matplotlib
# ----------------------------------------
#
# The following code visualizes the raw original data, the estimated block
# biases, and the normalized matrix using the CAIC method.
chromosomes = ["I", "II", "III", "IV", "V", "VI"]
fig, axes = plt.subplots(ncols=3, figsize=(14, 3))
axes[0].imshow(counts,
cmap="RdBu_r",