def test_load_cool(self):
"""Test loading of matrices in cool format"""
# Write a dummy bedgraph2 (basically a diagonal)
res, n_bins = 5000, 100000
chrom_names = ["c1", "c2", "c3"]
bins_per_chrom = [n_bins // 3, n_bins // 3, n_bins // 3 + n_bins % 3]
bins = pd.DataFrame(
{
"chrom": np.repeat(chrom_names, bins_per_chrom),
"start": range(0, res * (n_bins), res),
"end": range(res, res * (n_bins + 1), res),
}
)
pixels = pd.DataFrame(
{
"bin1_id": range(n_bins),
"bin2_id": range(n_bins),
"count": np.random.randint(0, 100, n_bins),
}
)
# Save dataframes into a cool file using cool API
cooler.create_cooler(self.tmp_path, bins, pixels)
# Load cool and check whether it was parsed correctly
mat, chroms, bins, bin_size = cio.load_cool(self.tmp_path)
# Median should work to estimate resolution id nbins >> nchroms
assert res == abs(int(np.nanmedian(bins.start.shift(1) - bins.start)))
assert res == bin_size
assert n_bins == bins.shape[0]
assert np.all(bins.columns == BIN_COLS)
assert np.all(chroms.columns == CHR_COLS)
assert mat.sum() == pixels["count"].sum()
import numpy as np
import unittest
from nose2.tools import params
import scipy.sparse as sp
import chromosight
import chromosight.utils
import chromosight.utils.preprocessing as preproc
import chromosight.utils.io as cio
mat, chroms, bins, res = cio.load_cool("data_test/example.cool")
mat = mat.tocsr()
# Get all intra-chromosomal matrices
intra_mats = [
mat[s:e, s:e] for s, e in zip(chroms["start_bin"], chroms["end_bin"])
]
class TestPreprocessing(unittest.TestCase):
def test_get_detectable_bins(self):
"""Check if correct bin indices are reported as detectable."""
# Make random matrix using uniform sample between 0 and 1
uniform_mat = sp.random(1000, 1000, density=0.1, format="csr")
# introduce outlier bin
uniform_mat[10, :] = 0.0
uniform_mat[:, 10] = 0.0
uniform_mat = uniform_mat.tocoo()
uniform_mat.eliminate_zeros()
det_bins = preproc.get_detectable_bins(uniform_mat,
inter=False,
n_mads=1)
# Check if symmetric mode return same detectable rows and cols