Merge pull request #90 from PixelgenTechnologies/feature/exe-1354

Feature/exe 1354

.gitattributes

@@ -1 +1,2 @@
 src/pixelator/report/webreport/template.html linguist-vendored
+tests/snapshots/test_plot/*.png filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -46,6 +46,7 @@ coverage.xml
 *.cover
 .hypothesis/
 .pytest_cache/
+tests/reports_mpl/*
 
 # Translations
 *.mo
@@ -113,3 +114,4 @@ ENV/
 pytest-results
 # pixelator file
 tests/data/ten_of_each.h5ad
+.DS_Store

CHANGELOG.md

@@ -9,6 +9,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+* Add QC plot showing UMIs per UPIA vs Tau.
+* Add plot functions showing edge rank and cell counts.
+* Add 2D and 3D graph plot functions.
+* Add plot functions showing colocalization and differential colocalization.
 * Performance improvements and reduced bundle size in QC report.
 * Improved console output in verbose mode.
 * Improved logging from multiprocessing jobs.

pyproject.toml

@@ -55,6 +55,7 @@ importlib-resources = "^5.12.0"
 fsspec = "^2023.12.2"
 fastparquet = "^2023.8.0"
 graspologic = "^3.3.0"
+plotly = "*"
 
 [tool.poetry.group.dev.dependencies]
 ruff = "*"
@@ -66,6 +67,7 @@ invoke = "*"
 isort = "*"
 pylint = "*"
 pytest = "^7.0.0"
+pytest-mpl = "^0.17.0"
 sphinx = "*"
 tox = "*"
 tox-current-env = "^0.0.11"
@@ -98,7 +100,9 @@ markers = [
     "integration_test: Marks a test as an integration test, which is often slow (deselect with '-m \"not integration_test\"')",
     "workflow_test: Marks a test as a complete pixelator workflow, which is extremely slow (deselect with '-m \"not workflow_test\"')"]
 
-addopts = ["-p no:pytest-workflow", "-m not workflow_test", "--benchmark-disable"]
+mpl-results-path = "tests/reports_mpl"
+mpl-generate-summary = "html"
+addopts = ["-p no:pytest-workflow", "-m not workflow_test", "--benchmark-disable", "--mpl"]
 filterwarnings = ["ignore::DeprecationWarning",]
 
 [tool.ruff]

src/pixelator/analysis/colocalization/__init__.py

@@ -29,10 +29,7 @@
     TransformationTypes,
 )
 from pixelator.graph.utils import Graph
-from pixelator.statistics import (
-    correct_pvalues,
-    log1p_transformation,
-)
+from pixelator.statistics import correct_pvalues, log1p_transformation
 
 logger = logging.getLogger(__name__)
 
@@ -288,3 +285,46 @@ def data():
 
     logger.debug("Colocalization scores for dataset computed")
     return scores
+
+
+def get_differential_colocalization(
+    colocalization_data_frame: pd.DataFrame,
+    target: str,
+    reference: str,
+    contrast_column: str = "sample",
+    use_z_score: bool = True,
+) -> pd.DataFrame:
+    """Calculate the differential colocalization.
+
+    :param colocalization_data_frame: The colocalization data frame.
+    :param target: The label for target components in the contrast_column.
+    :param reference: The label for reference components in the contrast_column.
+    :param contrast_column: The column to use for the contrast. Defaults to "sample".
+    :param use_z_score: Whether to use the z-score. Defaults to True.
+
+    :return: The differential colocalization.
+    :rtype: pd.DataFrame
+    """
+    if use_z_score:
+        value_column = "pearson_z"
+    else:
+        value_column = "pearson"
+
+    same_marker_mask = (
+        colocalization_data_frame["marker_1"] == colocalization_data_frame["marker_2"]
+    )
+    data_frame = colocalization_data_frame[~same_marker_mask]
+    colocalization_scores_source = data_frame[data_frame[contrast_column] == reference]
+    colocalization_scores_target = data_frame[data_frame[contrast_column] == target]
+    differential_colocalization = colocalization_scores_source.groupby(
+        ["marker_1", "marker_2"]
+    )[[value_column]].median().astype(float) - colocalization_scores_target.groupby(
+        ["marker_1", "marker_2"]
+    )[
+        [value_column]
+    ].median().astype(
+        float
+    )
+    differential_colocalization = differential_colocalization.reset_index()
+
+    return differential_colocalization

src/pixelator/graph/backends/implementations/_networkx.py

@@ -352,18 +352,16 @@ def _layout_coordinates(
         if layout_algorithm == "fruchterman_reingold_3d":
             layout_inst = nx.spring_layout(raw, dim=3, seed=random_seed)
         if layout_algorithm == "pmds":
-            return pmds_layout(raw, seed=random_seed, **kwargs)
+            layout_inst = pmds_layout(raw, seed=random_seed, **kwargs)
         if layout_algorithm == "pmds_3d":
-            return pmds_layout(raw, dim=3, seed=random_seed, **kwargs)
+            layout_inst = pmds_layout(raw, dim=3, seed=random_seed, **kwargs)
 
         coordinates = pd.DataFrame.from_dict(
             layout_inst,
             orient="index",
             columns=["x", "y"] if len(layout_inst[0]) == 2 else ["x", "y", "z"],
         )
-        coordinates.index = [
-            str(raw.nodes[node_idx]["name"]) for node_idx in layout_inst.keys()
-        ]
+
         return coordinates
 
     @staticmethod
@@ -425,7 +423,7 @@ def layout_coordinates(
             # Added here to avoid circular imports
             from pixelator.graph.utils import create_node_markers_counts
 
-            node_marker_counts = create_node_markers_counts(self)  # type: ignore
+            node_marker_counts = create_node_markers_counts(self, name_as_index=False)  # type: ignore
             df = pd.concat([coordinates, node_marker_counts], axis=1)
         else:
             df = coordinates
@@ -752,7 +750,8 @@ def pmds_layout(
     pivs = np.random.choice(g.nodes, pivots, replace=False)
 
     # Calculate the shortest path length from the pivots to all other nodes
-    A = nx.to_numpy_array(g, weight=None)
+    nodelist = list(g.nodes)
+    A = nx.to_numpy_array(g, weight=None, nodelist=nodelist)
     D = sp.sparse.csgraph.shortest_path(A, directed=False, unweighted=True)
     D_pivs = D[:, np.where(np.isin(g.nodes, pivs))[0]]
 
@@ -765,12 +764,7 @@ def pmds_layout(
     # in an abstract cartesian space
     _, _, Vh = np.linalg.svd(D_pivs_centered)
     coordinates = D_pivs_centered @ np.transpose(Vh)[:, :dim]
-    coordinates = pd.DataFrame(
-        coordinates, columns=["x", "y"] if dim == 2 else ["x", "y", "z"]
-    )
 
-    coordinates.index = [
-        str(g.nodes[node_idx]["name"]) for node_idx in coordinates.index
-    ]
+    coordinates = {nodelist[i]: coordinates[i, :] for i in range(coordinates.shape[0])}
 
     return coordinates

src/pixelator/graph/utils.py

@@ -148,6 +148,7 @@ def create_node_markers_counts(
     graph: Graph,
     k: int = 0,
     normalization: Optional[Literal["mean"]] = None,
+    name_as_index: bool = True,
 ) -> pd.DataFrame:
     """Create a matrix of marker counts for each in the graph.
 
@@ -180,16 +181,17 @@ def create_node_markers_counts(
     node_marker_counts = pd.DataFrame.from_records(
         list(graph.vs.get_attribute("markers")),
         columns=markers,
-        index=list(graph.vs.get_attribute("name")),
+        index=list(graph.raw.nodes),
     )
     node_marker_counts = node_marker_counts.reindex(
         sorted(node_marker_counts.columns), axis=1
     )
     node_marker_counts.columns.name = "markers"
     node_marker_counts.columns = node_marker_counts.columns.astype("string[pyarrow]")
-    node_marker_counts.index = pd.Index(
-        list(graph.vs.get_attribute("name")), dtype="string[pyarrow]", name="node"
-    )
+    if name_as_index:
+        node_marker_counts.index = pd.Index(
+            list(graph.vs.get_attribute("name")), dtype="string[pyarrow]", name="node"
+        )
     if k == 0:
         return node_marker_counts