def per_cell_normalize(adata, results_folder):
# get start time
start = time()
# normalize per cell
# already normalize BEFORE saving "raw" - as recommended in the scanpy tutorial
normalize_per_cell(adata, counts_per_cell_after=1e4)
print('adata normalized per cell')
# keep raw copy
adata.raw = log1p(adata, copy=True)
print('log1p values saved into adata.raw')
# make log entries
logging.info('Per cell normalization completed successfully.')
logging.info("\tTime for per-cell normalization: " +
str(round(time() - start, 3)) + 's')
# export to file
start = time()
export_cp10k(adata, basepath=results_folder)
logging.info('cp10k values exported to file.')
logging.info("\tTime for cp10k export: " + str(round(time() - start, 3)) +
's')
return (adata)
def test(self, Xtest):
testdata = AnnData(Xtest)
normalize_per_cell(testdata, 1000, min_counts=0)
log1p(testdata)
testdata.X = _toarray(testdata.X)
dxixk = scipy.spatial.distance.cdist(testdata.X, self.xkibar)
return dxixk.argmin(axis=1)
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 train(self, adata):
adata = adata.copy()
adata.X = _toarray(adata.X)
normalize_per_cell(adata, 1000, min_counts=0)
log1p(adata)
adata = process_clusts(adata)
self.xkibar = np.array(
[
adata.X[adata.uns["clusterindices"][k]].mean(axis=0).tolist()
for k in range(adata.uns["num_clusts"])
]
)