def test_concatenate_two_expression_matrices():
a_data = np.array([[0, 1], [1, 0]])
b_data = np.array([[0], [1]])
dims = [Features.CELLS, Features.GENES]
a_coords = [(Features.CELLS, [0, 1]), (Features.GENES, ["x", "y"])]
b_coords = [(Features.CELLS, [0, 1]), (Features.GENES, ["x"])]
a = ExpressionMatrix(a_data, dims=dims, coords=a_coords)
b = ExpressionMatrix(b_data, dims=dims, coords=b_coords)
concatenated = concatenate([a, b])
expected = np.array([[0, 1], [1, 0], [0, np.nan], [1, np.nan]])
np.testing.assert_equal(concatenated.values, expected)
def to_expression_matrix(self) -> ExpressionMatrix:
"""
Generates a cell x gene count matrix where each cell is annotated with spatial metadata.
Requires that spots in the IntensityTable have been assigned to cells.
Returns
-------
ExpressionMatrix :
cell x gene expression table
"""
if Features.CELL_ID not in self.coords:
raise KeyError(
"IntensityTable must have 'cell_id' assignments for each cell before this function "
"can be called. See starfish.spots.AssignTargets.Label.")
grouped = self.to_features_dataframe().groupby(
[Features.CELL_ID, Features.TARGET])
counts = grouped.count().iloc[:, 0].unstack().fillna(0)
# rename unassigned spots
counts.rename(index={'nan': 'unassigned'}, inplace=True)
# remove and store 'nan' target counts
nan_target_counts = np.zeros(counts.shape[0])
if 'nan' in counts.columns:
nan_target_counts = counts['nan'].values
counts.drop(columns='nan', inplace=True)
if self.has_physical_coords:
grouped = self.to_features_dataframe().groupby(
[Features.CELL_ID])[[
Axes.X.value, Axes.Y.value, Axes.ZPLANE.value,
Coordinates.X.value, Coordinates.Y.value,
Coordinates.Z.value
]]
else:
grouped = self.to_features_dataframe().groupby([
Features.CELL_ID
])[[Axes.X.value, Axes.Y.value, Axes.ZPLANE.value]]
min_ = grouped.min()
max_ = grouped.max()
coordinate_df = min_ + (max_ - min_) / 2
metadata = {
name: (Features.CELLS, data.values)
for name, data in coordinate_df.items()
}
metadata[Features.AREA] = (Features.CELLS,
np.full(counts.shape[0], fill_value=np.nan))
metadata["number_of_undecoded_spots"] = (Features.CELLS,
nan_target_counts)
# add genes to the metadata
metadata.update({Features.GENES: counts.columns.values})
metadata.update(
{Features.CELL_ID: (Features.CELLS, counts.index.values)})
mat = ExpressionMatrix(data=counts.values,
dims=(Features.CELLS, Features.GENES),
coords=metadata,
name='expression_matrix')
return mat