def test_filter_genes(self, adata, adata_dist):
filter_genes(adata_dist, min_cells=2)
result = materialize_as_ndarray(adata_dist.X)
filter_genes(adata, min_cells=2)
assert result.shape == adata.shape
assert result.shape == (adata.n_obs, adata.n_vars)
npt.assert_allclose(result, adata.X)
def test_normalize_per_cell(self, adata, adata_dist):
normalize_per_cell(adata_dist)
result = materialize_as_ndarray(adata_dist.X)
normalize_per_cell(adata)
assert result.shape == adata.shape
assert result.shape == (adata.n_obs, adata.n_vars)
npt.assert_allclose(result, adata.X)
def test_log1p(self, adata, adata_dist):
log1p(adata_dist)
result = materialize_as_ndarray(adata_dist.X)
log1p(adata)
assert result.shape == adata.shape
assert result.shape == (adata.n_obs, adata.n_vars)
npt.assert_allclose(result, adata.X)
def test_scale(self, adata, adata_dist):
if isinstance(adata_dist.X, da.Array):
return # fails for dask
scale(adata_dist)
result = materialize_as_ndarray(adata_dist.X)
scale(adata)
assert result.shape == adata.shape
assert result.shape == (adata.n_obs, adata.n_vars)
npt.assert_allclose(result, adata.X)
def test_recipe_zheng17(self, adata, adata_dist):
recipe_zheng17(adata_dist, n_top_genes=100)
result = materialize_as_ndarray(adata_dist.X)
recipe_zheng17(adata, n_top_genes=100)
assert result.shape == adata.shape
assert result.shape == (adata.n_obs, adata.n_vars)
# Note the low tolerance required to get this to pass.
# Not sure why results diverge so much. (Seems to be scaling again.)
# Find the element that differs the most with
# import numpy
# am = (numpy.absolute(result - adata.X)/ numpy.absolute(adata.X)).argmax()
# ind = numpy.unravel_index(am, result.shape)
# print(result[ind], adata.X[ind])
npt.assert_allclose(result, adata.X, 1e-1)