Parameterize standardization

inferelator_velocity/program_graph.py

@@ -18,6 +18,7 @@
 def global_graph(
     data,
     layer="X",
+    standardize_layer_data=True,
     neighbors=None,
     npcs=None,
     use_sparse=True,
@@ -31,6 +32,8 @@ def global_graph(
     :type data: ad.AnnData
     :param layer: Layer to use for kNN, defaults to "X"
     :type layer: str, optional
+    :param standardize_layer_data: Depth-normalize and log1p data
+    :type standardize_layer_data: bool, optional
     :param neighbors: Search space for k neighbors,
         defaults to 15 - 105 by 10
     :type neighbors: np.ndarray, optional
@@ -63,7 +66,8 @@ def global_graph(
         npcs=npcs,
         verbose=verbose,
         use_sparse=use_sparse,
-        connectivity=connectivity
+        connectivity=connectivity,
+        standardize_count_data=standardize_layer_data
     )
 
     vprint(

inferelator_velocity/utils/noise2self.py

@@ -43,7 +43,8 @@ def knn_noise2self(
     metric='euclidean',
     return_errors=False,
     use_sparse=True,
-    connectivity=False
+    connectivity=False,
+    standardize_count_data=True
 ):
     """
     Select an optimal set of graph parameters based on noise2self
@@ -100,8 +101,10 @@ def knn_noise2self(
 
     data_obj = ad.AnnData(count_data.astype(np.float32))
 
-    sc.pp.normalize_per_cell(data_obj)
-    sc.pp.log1p(data_obj)
+    if standardize_count_data:
+        sc.pp.normalize_per_cell(data_obj)
+        sc.pp.log1p(data_obj)
+
     sc.pp.pca(data_obj, n_comps=np.max(npcs), zero_center=True)
 
     mses = np.zeros((len(npcs), len(neighbors)))

inferelator_velocity/velocity.py

@@ -27,7 +27,7 @@ def calc_velocity(
                 (n_idx == i).nonzero()[0][0],
                 wrap_time=wrap_time
             )
-            for i, n_idx in _find_local(expr, neighbor_graph)
+            for i, n_idx in _find_local(neighbor_graph)
         ]
     )
 
@@ -93,18 +93,18 @@ def _calc_local_velocity(
     return _np.dot(x_for_hat, y_diff)
 
 
-def _find_local(expr, neighbor_graph):
+def _find_local(neighbor_graph):
     """
-    Find a return an expression matrix for a locally connected graph
+    Find a return indices from a locally connected graph
 
-    :param expr: Samples x Genes numpy or scipy with expression data
-    :type expr: np.ndarray, sp.sparse.csr_matrix
     :param neighbor_graph: Samples x Samples connectivity matrix,
         where any non-zero value is connected.
-    :return:
+    :type neighbor_graph: np.ndarray, sp.sparse.spmatrix
+    :yield: Center index, connected indices
+    :rtype: int, np.ndarray
     """
 
-    n, m = expr.shape
+    n, _ = neighbor_graph.shape
     neighbor_sparse = _is_sparse(neighbor_graph)
 
     for i in trange(n):