def timerange_filter(df: pd.DataFrame, timeranges: np.ndarray) -> pd.DataFrame:
"""\
Returns the filter that can be used to get all the rows of dataframe df where
timestamp column falls in timeranges, a 2D NumPy array with start time in the
first column and end time in the second column; both in miliseconds
"""
timeranges = pd.DataFrame(timeranges, columns=['from', 'to'])
inds = timeranges.apply(lambda x: in_range(x, df), axis=1)
return inds.T.any(axis=1)
def lr_pandas(
spot_lr1: np.ndarray,
spot_lr2: np.ndarray,
neighbours: list,
) -> np.ndarray:
"""Calculate the lr scores for each spot.
Parameters
----------
spot_lr1: pd.DataFrame Cells*Ligands
spot_lr2: pd.DataFrame Cells*Receptors
neighbours: list List of neighbours by indices for each spot.
Returns
-------
lr_scores: numpy.ndarray Cells*LR-scores.
"""
# function to calculate mean of lr2 expression between neighbours or within spot (distance==0) for each spot
def mean_lr2(x):
# get lr2 expressions from the neighbour(s)
n_spots = neighbours[spot_lr2.index.tolist().index(x.name)]
nbs = spot_lr2.loc[n_spots, :]
if nbs.shape[0] > 0: # if neighbour exists
return nbs.sum() / nbs.shape[0]
else:
return 0
# mean of lr2 expressions from neighbours of each spot
nb_lr2 = spot_lr2.apply(mean_lr2, axis=1)
# check whether neighbours exist
try:
nb_lr2.shape[1]
except:
raise ValueError("No neighbours found within given distance.")
# keep value of nb_lr2 only when lr1 is also expressed on the spots
spot_lr = pd.DataFrame(
spot_lr1.values * (nb_lr2.values > 0) +
(spot_lr1.values > 0) * nb_lr2.values, ).sum(axis=1)
return spot_lr.values / 2
