scib-metrics 0.5.3__tar.gz → 0.5.4__tar.gz

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/.pre-commit-config.yaml

@@ -11,7 +11,7 @@ repos:
     hooks:
       - id: prettier
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.9.6
+    rev: v0.11.6
     hooks:
       - id: ruff
         types_or: [python, pyi, jupyter]

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/CHANGELOG.md

@@ -10,6 +10,18 @@ and this project adheres to [Semantic Versioning][].
 
 ## 0.6.0 (unreleased)
 
+## 0.5.4 (2025-04-23)
+
+### Fixed
+
+- Apply default values for benchmarker metrics {pr}`203`.
+
+## 0.5.3 (2025-02-17)
+
+### Removed
+
+- Reverted a change that was needed for scib-autotune in scvi-tools {pr}`189`.
+
 ## 0.5.2 (2025-02-13)
 
 ### Added

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scib-metrics
-Version: 0.5.3
+Version: 0.5.4
 Summary: Accelerated and Python-only scIB metrics
 Project-URL: Documentation, https://scib-metrics.readthedocs.io/
 Project-URL: Source, https://github.com/yoseflab/scib-metrics

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/pyproject.toml

@@ -5,7 +5,7 @@ requires = ["hatchling"]
 
 [project]
 name = "scib-metrics"
-version = "0.5.3"
+version = "0.5.4"
 description = "Accelerated and Python-only scIB metrics"
 readme = "README.md"
 requires-python = ">=3.10"

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/src/scib_metrics/benchmark/_core.py

@@ -136,8 +136,8 @@ class Benchmarker:
         batch_key: str,
         label_key: str,
         embedding_obsm_keys: list[str],
-        bio_conservation_metrics: BioConservation | None,
-        batch_correction_metrics: BatchCorrection | None,
+        bio_conservation_metrics: BioConservation | None = BioConservation(),
+        batch_correction_metrics: BatchCorrection | None = BatchCorrection(),
         pre_integrated_embedding_obsm_key: str | None = None,
         n_jobs: int = 1,
         progress_bar: bool = True,

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/src/scib_metrics/metrics/_kbet.py

@@ -138,8 +138,14 @@ def kbet_per_label(
     conn_graph = X.knn_graph_connectivities
 
     # prepare call of kBET per cluster
-    kbet_scores = {"cluster": [], "kBET": []}
-    for clus in np.unique(labels):
+    clusters = []
+    clusters, counts = np.unique(labels, return_counts=True)
+    skipped = clusters[counts > 10]
+    clusters = clusters[counts <= 10]
+    kbet_scores = {"cluster": list(skipped), "kBET": [np.nan] * len(skipped)}
+    logger.info(f"{len(skipped)} clusters consist of a single batch or are too small. Skip.")
+
+    for clus in clusters:
         # subset by label
         mask = labels == clus
         conn_graph_sub = conn_graph[mask, :][:, mask]
@@ -147,60 +153,55 @@ def kbet_per_label(
         n_obs = conn_graph_sub.shape[0]
         batches_sub = batches[mask]
 
-        # check if neighborhood size too small or only one batch in subset
-        if np.logical_or(n_obs < 10, len(np.unique(batches_sub)) == 1):
-            logger.info(f"{clus} consists of a single batch or is too small. Skip.")
-            score = np.nan
-        else:
-            quarter_mean = np.floor(np.mean(pd.Series(batches_sub).value_counts()) / 4).astype("int")
-            k0 = np.min([70, np.max([10, quarter_mean])])
-            # check k0 for reasonability
-            if k0 * n_obs >= size_max:
-                k0 = np.floor(size_max / n_obs).astype("int")
+        quarter_mean = np.floor(np.mean(pd.Series(batches_sub).value_counts()) / 4).astype("int")
+        k0 = np.min([70, np.max([10, quarter_mean])])
+        # check k0 for reasonability
+        if k0 * n_obs >= size_max:
+            k0 = np.floor(size_max / n_obs).astype("int")
+
+        n_comp, labs = scipy.sparse.csgraph.connected_components(conn_graph_sub, connection="strong")
+
+        if n_comp == 1:  # a single component to compute kBET on
+            try:
+                diffusion_n_comps = np.min([diffusion_n_comps, n_obs - 1])
+                nn_graph_sub = diffusion_nn(conn_graph_sub, k=k0, n_comps=diffusion_n_comps)
+                # call kBET
+                score, _, _ = kbet(
+                    nn_graph_sub,
+                    batches=batches_sub,
+                    alpha=alpha,
+                )
+            except ValueError:
+                logger.info("Diffusion distance failed. Skip.")
+                score = 0  # i.e. 100% rejection
 
-            n_comp, labs = scipy.sparse.csgraph.connected_components(conn_graph_sub, connection="strong")
+        else:
+            # check the number of components where kBET can be computed upon
+            comp_size = pd.Series(labs).value_counts()
+            # check which components are small
+            comp_size_thresh = 3 * k0
+            idx_nonan = np.flatnonzero(np.in1d(labs, comp_size[comp_size >= comp_size_thresh].index))
+
+            # check if 75% of all cells can be used for kBET run
+            if len(idx_nonan) / len(labs) >= 0.75:
+                # create another subset of components, assume they are not visited in a diffusion process
+                conn_graph_sub_sub = conn_graph_sub[idx_nonan, :][:, idx_nonan]
+                conn_graph_sub_sub.sort_indices()
 
-            if n_comp == 1:  # a single component to compute kBET on
                 try:
-                    diffusion_n_comps = np.min([diffusion_n_comps, n_obs - 1])
-                    nn_graph_sub = diffusion_nn(conn_graph_sub, k=k0, n_comps=diffusion_n_comps)
+                    diffusion_n_comps = np.min([diffusion_n_comps, conn_graph_sub_sub.shape[0] - 1])
+                    nn_results_sub_sub = diffusion_nn(conn_graph_sub_sub, k=k0, n_comps=diffusion_n_comps)
                     # call kBET
                     score, _, _ = kbet(
-                        nn_graph_sub,
-                        batches=batches_sub,
+                        nn_results_sub_sub,
+                        batches=batches_sub[idx_nonan],
                         alpha=alpha,
                     )
                 except ValueError:
                     logger.info("Diffusion distance failed. Skip.")
                     score = 0  # i.e. 100% rejection
-
-            else:
-                # check the number of components where kBET can be computed upon
-                comp_size = pd.Series(labs).value_counts()
-                # check which components are small
-                comp_size_thresh = 3 * k0
-                idx_nonan = np.flatnonzero(np.in1d(labs, comp_size[comp_size >= comp_size_thresh].index))
-
-                # check if 75% of all cells can be used for kBET run
-                if len(idx_nonan) / len(labs) >= 0.75:
-                    # create another subset of components, assume they are not visited in a diffusion process
-                    conn_graph_sub_sub = conn_graph_sub[idx_nonan, :][:, idx_nonan]
-                    conn_graph_sub_sub.sort_indices()
-
-                    try:
-                        diffusion_n_comps = np.min([diffusion_n_comps, conn_graph_sub_sub.shape[0] - 1])
-                        nn_results_sub_sub = diffusion_nn(conn_graph_sub_sub, k=k0, n_comps=diffusion_n_comps)
-                        # call kBET
-                        score, _, _ = kbet(
-                            nn_results_sub_sub,
-                            batches=batches_sub[idx_nonan],
-                            alpha=alpha,
-                        )
-                    except ValueError:
-                        logger.info("Diffusion distance failed. Skip.")
-                        score = 0  # i.e. 100% rejection
-                else:  # if there are too many too small connected components, set kBET score to 0
-                    score = 0  # i.e. 100% rejection
+            else:  # if there are too many too small connected components, set kBET score to 0
+                score = 0  # i.e. 100% rejection
 
         kbet_scores["cluster"].append(clus)
         kbet_scores["kBET"].append(score)

{scib_metrics-0.5.3 → scib_metrics-0.5.4}/tests/test_benchmarker.py

@@ -21,6 +21,20 @@ def test_benchmarker():
     bm.plot_results_table()
 
 
+def test_benchmarker_default():
+    ad, emb_keys, batch_key, labels_key = dummy_benchmarker_adata()
+    bm = Benchmarker(
+        ad,
+        batch_key,
+        labels_key,
+        emb_keys,
+    )
+    bm.benchmark()
+    results = bm.get_results()
+    assert isinstance(results, pd.DataFrame)
+    bm.plot_results_table()
+
+
 def test_benchmarker_custom_metric_booleans():
     bioc = BioConservation(
         isolated_labels=False, nmi_ari_cluster_labels_leiden=False, silhouette_label=False, clisi_knn=True