def test_batch_mean_center():
X = data.generate_positive_sparse_matrix()
sample_idx = np.random.choice([0, 1], X.shape[0], replace=True)
X[sample_idx == 1] += 1
Y = X.copy()
Y[sample_idx == 0] -= np.mean(Y[sample_idx == 0], axis=0)[None, :]
Y[sample_idx == 1] -= np.mean(Y[sample_idx == 1], axis=0)[None, :]
utils.assert_all_close(np.mean(Y[sample_idx == 0], axis=0), 0)
utils.assert_all_close(np.mean(Y[sample_idx == 1], axis=0), 0)
matrix.test_dense_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=partial(
scprep.normalize.batch_mean_center,
sample_idx=sample_idx))
matrix.test_sparse_matrix_types(
X, utils.assert_transform_raises,
transform=partial(
scprep.normalize.batch_mean_center,
sample_idx=sample_idx),
exception=ValueError)
X = data.generate_positive_sparse_matrix()
Y = X.copy()
Y -= np.mean(Y, axis=0)[None, :]
matrix.test_dense_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=partial(
scprep.normalize.batch_mean_center))
matrix.test_sparse_matrix_types(
X, utils.assert_transform_raises,
transform=partial(
scprep.normalize.batch_mean_center),
exception=ValueError)
def test_eps_too_low(self):
utils.assert_all_close(
self.Y_random,
scprep.reduce.pca(self.X_sparse,
n_components=100,
eps=0.0001,
seed=42),
)
def test_libsize_norm_rescale_mean(self):
Y = self.X_norm * self.mean
utils.assert_all_close(Y.sum(1), np.mean(np.sum(self.X, 1)))
matrix.test_all_matrix_types(
self.X,
utils.assert_transform_equivalent,
Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close,
rescale='mean')
def test_singular_values_sparse(self):
utils.assert_all_close(
self.S_full,
scprep.reduce.pca(
self.X_sparse,
n_components=50,
eps=0.3,
seed=42,
return_singular_values=True,
)[1],
atol=1e-14,
)
def test_libsize_norm_rescale_median(self):
Y = self.X_norm * self.median
utils.assert_all_close(Y.sum(1), np.median(np.sum(self.X, 1)))
Y2, libsize2 = scprep.normalize.library_size_normalize(
self.X, return_library_size=True)
np.testing.assert_allclose(Y, Y2)
np.testing.assert_allclose(self.libsize, libsize2)
matrix.test_all_matrix_types(
self.X,
utils.assert_transform_equivalent,
Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close)
def test_batch_mean_center(self):
X = self.X.copy()
X[self.sample_idx == 1] += 1
Y = X.copy()
Y[self.sample_idx == 0] -= np.mean(Y[self.sample_idx == 0],
axis=0)[None, :]
Y[self.sample_idx == 1] -= np.mean(Y[self.sample_idx == 1],
axis=0)[None, :]
utils.assert_all_close(np.mean(Y[self.sample_idx == 0], axis=0), 0)
utils.assert_all_close(np.mean(Y[self.sample_idx == 1], axis=0), 0)
matrix.test_dense_matrix_types(
X,
utils.assert_transform_equivalent,
Y=Y,
transform=scprep.normalize.batch_mean_center,
sample_idx=self.sample_idx)
def test_libsize_norm():
X = data.generate_positive_sparse_matrix()
median = np.median(X.sum(axis=1))
Y = normalize(X, 'l1') * median
utils.assert_all_close(Y.sum(1), np.median(np.sum(X, 1)))
matrix.test_all_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close)
mean = np.mean(X.sum(axis=1))
X = data.generate_positive_sparse_matrix()
Y = normalize(X, 'l1') * mean
utils.assert_all_close(Y.sum(1), np.mean(np.sum(X, 1)))
matrix.test_all_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close, rescale='mean')
X = data.generate_positive_sparse_matrix()
Y = normalize(X, 'l1')
matrix.test_all_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close, rescale=None)
matrix.test_all_matrix_types(
X, utils.assert_transform_equivalent, Y=Y,
transform=scprep.normalize.library_size_normalize,
check=utils.assert_all_close, rescale=1)
assert_raise_message(
ValueError,
"Expected rescale in ['median', 'mean'], a number or `None`. "
"Got invalid",
scprep.normalize.library_size_normalize,
X, rescale='invalid')
X[:X.shape[0] // 2 + 1] = 0
assert_warns_message(
UserWarning,
"Median library size is zero. "
"Rescaling to mean instead.",
scprep.normalize.library_size_normalize,
X, rescale='median')