Bug fix

simulation/002_zone2counts.py

@@ -33,7 +33,7 @@ def main():
 					 type = str,
 					 default = None,
 					 help = 'Output directory. Default: None to use the same directory as the input.')
-	prs.add_argument('-oa','--output-anndata', action=argparse.BooleanOptionalAction, default=True,
+	prs.add_argument('-oa','--output-anndata', action='store_true',
 					 help = 'Whether to output entire anndata objects for synthetic spatial data ' \
 							'and paired single cell data. Default: True.' )				
 

smoother/losses.py

@@ -308,7 +308,7 @@ def calc_corr_decay_stats(self, coords, min_k = 0, max_k = 50, cov_ind = 0, retu
 			try:
 				cov = torch.cholesky_inverse(torch.linalg.cholesky(self.inv_cov[cov_ind].to_dense()))
 			except RuntimeError as exc:
-				raise RuntimeError(f"The current loss ({self.prior}, l={self.rho}) "
+				raise RuntimeError(f"The current loss ({self.prior}, rho={self.rho}) "
 									"contains an improper spatial covariance structure. "
 									"Please use a different spatial prior or scale "
 									"if you intend to calculate covariance decay.") from exc

smoother/simulation/utils.py

@@ -11,7 +11,7 @@
 from scipy.sparse import csr_matrix,lil_matrix
 import anndata
 from tqdm import tqdm
-from smoother.weights import _coordinate_to_weights_knn
+from smoother.weights import coordinate_to_weights_knn_dense
 
 
 def sample_cell_indices(generation_snrna, annot_label, cell_count_df, cell_capture_eff_df):
@@ -66,7 +66,7 @@ def smooth_by_neighbors(shared_prop, locations2cells_matrix, coords, n_experimen
 		smoothed cell index matrix (n_spots x n_cells).
 	'''
 	# compute adjacency matrix
-	swm = _coordinate_to_weights_knn(coords, k=4, row_scale=True) * shared_prop
+	swm = coordinate_to_weights_knn_dense(coords, k=4, row_scale=True) * shared_prop
 	swm = swm.fill_diagonal_(1 - shared_prop)
 
 	# compute smoothed cell index matrix

smoother/utils.py

@@ -41,12 +41,19 @@ def normalize_minmax(x, min_zero = True, return_scale = False):
 
 	return x_norm
 
-def _pca(features, dim):
-	"""Dimension reduction by PCA."""
+def _z_score(features):
+	"""Z-score standardization."""
 	# standardize and remove constant features
+	# features.shape: num_feature x num_spot
 	features_var = features.std(1) # feature level variance
 	features_scaled = (features - features.mean(1, keepdim=True)) / features_var[:, None]
 	features_scaled = features_scaled[features_var > 0,:]
+	return features_scaled
+
+def _pca(features, dim):
+	"""Dimension reduction by PCA."""
+	# standardize and remove constant features
+	features_scaled = _z_score(features)
 
 	# run pca
 	torch.manual_seed(0) # for repeatability, fix the random seed for pca_lowrank