drop leidenalg dependency

Author: katosh

.gitignore

@@ -10,3 +10,4 @@ notebooks/testing.ipynb
 dist/
 .vscode/
 data/
+*.ipynb

README.md

@@ -75,6 +75,7 @@ Release Notes
  ### Version 1.4.1
   * update `LICENSE` file to be consistent with MIT - license
   * implement `plot_trajectories` to show multiple poaths on the UMAP
+  * Drop leiden dependency to allow Python >= 3.13. igrph is used instead.
 
  ### Version 1.4.0
  * Made pygam an optional dependency that can be installed with `pip install palantir[gam]` or `pip install palantir[full]`

pyproject.toml

@@ -34,7 +34,6 @@ dependencies = [
     "scikit-learn",
     "joblib",
     "fcsparser>=0.1.2",
-    "leidenalg>=0.9.1",
     "matplotlib>=3.8.0",
     "anndata>=0.8.0",
     "scanpy>=1.6.0",
@@ -103,4 +102,4 @@ exclude_lines = [
     "raise ImportError",
     "def _return_cell",
     "print"
-]
+]

src/palantir/io.py

@@ -191,7 +191,8 @@ def from_fcs(
     """
     # Parse the fcs file
     text, data = fcsparser.parse(fcs_file)
-    data = data.astype(np.float64)
+    # Use view instead of newbyteorder for NumPy 2.0 compatibility
+    data = data.astype(np.float64, copy=False)
 
     # Extract the S and N features (Indexing assumed to start from 1)
     # Assumes channel names are in S

src/palantir/plot.py

@@ -1984,7 +1984,7 @@ def plot_gene_trend_clusters(
     )
 
     # Obtain unique clusters and prepare figure
-    if pd.api.types.is_categorical_dtype(clusters):
+    if isinstance(clusters.dtype, pd.CategoricalDtype):
         cluster_labels = clusters.cat.categories
     else:
         # Filter out NaN values to avoid issues with np.NaN vs np.nan
@@ -2096,7 +2096,9 @@ def gene_score_histogram(
             label=f"{quantile:.0%} percentile",
         )
 
-    ax.legend()
+    # Only create legend if we have elements with labels
+    if quantile is not None:
+        ax.legend()
     ax.set_xlabel(f"{score_key} score")
     ax.set_ylabel("# of genes")
 

src/palantir/presults.py

@@ -504,7 +504,13 @@ def cluster_gene_trends(
 
     gt_ad = AnnData(trends.values, dtype=np.float32)
     sc.pp.neighbors(gt_ad, n_neighbors=n_neighbors, use_rep="X")
-    sc.tl.leiden(gt_ad, **kwargs)
+    
+    # Add required kwargs for leiden with igraph backend to avoid FutureWarning
+    leiden_kwargs = {'flavor': 'igraph', 'n_iterations': 2, 'directed': False}
+    # Override with user-provided kwargs if any
+    leiden_kwargs.update(kwargs)
+    
+    sc.tl.leiden(gt_ad, **leiden_kwargs)
 
     communities = pd.Series(gt_ad.obs["leiden"].values, index=trends.index)
 

tests/test_plot.py

@@ -401,14 +401,19 @@ def test_plot_palantir_results_custom_args(mock_anndata):
 
 # Test with AnnData and all keys available
 def test_plot_terminal_state_probs_anndata(mock_anndata, mock_cells):
+    # Close existing figures to avoid warnings
+    plt.close('all')
     fig = plot_terminal_state_probs(mock_anndata, mock_cells)
     assert isinstance(fig, plt.Figure)
+    plt.close(fig)
 
 
 # Test with DataFrame and PResults
 def test_plot_terminal_state_probs_dataframe(mock_data, mock_presults, mock_cells):
+    plt.close('all')
     fig = plot_terminal_state_probs(mock_data, mock_cells, pr_res=mock_presults)
     assert isinstance(fig, plt.Figure)
+    plt.close(fig)
 
 
 # Test ValueError for missing pr_res in DataFrame input
@@ -616,31 +621,42 @@ def test_plot_gene_trend_heatmaps(mock_anndata):
 
 
 def test_plot_gene_trend_clusters(mock_anndata):
+    # Close all figures to start clean
+    plt.close('all')
+    
     # Test with AnnData object
     fig = plot_gene_trend_clusters(mock_anndata, branch_name="a", clusters="clusters")
     assert isinstance(fig, plt.Figure)
 
     # We expect one subplot per unique non-NaN cluster
     unique_clusters = [x for x in mock_anndata.var["clusters"].unique() if not pd.isna(x)]
     assert len(fig.axes) == len(unique_clusters)
+    plt.close(fig)
 
+    # Close all figures again before next test
+    plt.close('all')
+    
     # Test DataFrame input
     trends_df = mock_anndata.varm["gene_trends_a"]
     clusters_series = mock_anndata.var["clusters"]
     fig_df = plot_gene_trend_clusters(trends_df, clusters=clusters_series)
 
     assert isinstance(fig_df, plt.Figure)
     assert len(fig_df.axes) == len(unique_clusters)
-
-    plt.close(fig)
+    
     plt.close(fig_df)
+    plt.close('all')
 
 
 def test_gene_score_histogram(mock_anndata):
+    # Close all figures before testing
+    plt.close('all')
+    
     # Test with minimum required parameters
     fig = gene_score_histogram(mock_anndata, "gene_score")
     assert isinstance(fig, plt.Figure)
     plt.close(fig)
+    plt.close('all')
 
     # Test with optional parameters
     fig = gene_score_histogram(
@@ -652,15 +668,18 @@ def test_gene_score_histogram(mock_anndata):
     )
     assert isinstance(fig, plt.Figure)
     plt.close(fig)
+    plt.close('all')
 
-    # Test with None quantile
+    # Test with None quantile - this would previously cause a warning
+    # about empty legend when quantile is None
     fig = gene_score_histogram(
         mock_anndata,
         "gene_score",
         quantile=None,
     )
     assert isinstance(fig, plt.Figure)
     plt.close(fig)
+    plt.close('all')
 
 
 def test_gene_score_histogram_errors(mock_anndata):