feat: Include popV version in model metadata and refine metadata field handling and tag formatting. Pin broken dependencies.

Author: canergen

.pre-commit-config.yaml

@@ -2,8 +2,7 @@ fail_fast: false
 default_language_version:
   python: python3
 default_stages:
-  - pre-commit
-  - pre-push
+  - commit
 minimum_pre_commit_version: 2.16.0
 repos:
   - repo: https://github.com/asottile/blacken-docs
@@ -18,7 +17,7 @@ repos:
         types: [yaml]
 
   - repo: https://github.com/executablebooks/mdformat
-    rev: 0.7.22
+    rev: 1.0.0
     hooks:
       - id: mdformat
         additional_dependencies:
@@ -31,7 +30,7 @@ repos:
           )$
 
   - repo: https://github.com/igorshubovych/markdownlint-cli
-    rev: v0.45.0
+    rev: v0.47.0
     hooks:
       - id: markdownlint-fix
         exclude: |
@@ -41,7 +40,7 @@ repos:
           )$
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.13.1
+    rev: v0.15.4
     hooks:
       - id: ruff
         args: [--fix, --exit-non-zero-on-fix]

popv/algorithms/_bbknn.py

@@ -91,13 +91,20 @@ def compute_integration(self, adata):
         logging.info("Integrating data with bbknn")
         if len(adata.obs[self.batch_key].unique()) > 100:
             self.method_kwargs["neighbors_within_batch"] = 1
-        if settings.cuml:
+        if len(adata.obs[self.batch_key].unique()) > 200 and settings.cuml:
+            logging.warning(
+                f"Number of batches is {len(adata.obs[self.batch_key].unique())}, skipping RAPIDS BBKNN and running on CPU."
+            )
+            cuml = False
+        else:
+            cuml = settings.cuml
+        if cuml:
             import rapids_singlecell as rsc
 
             self.method_kwargs.pop("approx", None)  # approx not supported in rsc
             self.method_kwargs.pop("use_annoy", None)  # use_annoy not supported in rsc
             rsc.pp.bbknn(
-                adata, batch_key=self.batch_key, use_rep="X_pca", algorithm="ivfflat", **self.method_kwargs, trim=0
+                adata, batch_key=self.batch_key, use_rep="X_pca", algorithm="ivfflat", **self.method_kwargs
             )
         else:
             sc.external.pp.bbknn(adata, batch_key=self.batch_key, use_rep="X_pca", **self.method_kwargs)

popv/algorithms/_onclass.py

@@ -188,6 +188,8 @@ def predict(self, adata):
                 else:
                     adata.obs[col] = adata.uns["unknown_celltype_label"]
                     adata.obs[col] = adata.obs[col].astype(str)  # Set dtype to string
+        adata.obs[self.result_key] = adata.obs[self.result_key].astype(str)
+        adata.obs[self.seen_result_key] = adata.obs[self.seen_result_key].astype(str)
         adata.obs.loc[adata.obs["_predict_cells"] == "relabel", result_df.columns] = result_df
         if self.return_probabilities:
             if f"{self.result_key}_probabilities" not in adata.obsm:

popv/algorithms/_scanvi.py

@@ -208,7 +208,7 @@ def compute_umap(self, adata):
             Anndata object. Results are stored in adata.obsm[self.umap_key].
         """
         if self.compute_umap_embedding:
-            logging.info(f'Saving UMAP of BBKNN results to adata.obsm["{self.umap_key}"]')
+            logging.info(f'Saving UMAP of scANVI results to adata.obsm["{self.umap_key}"]')
             if settings.cuml:
                 import rapids_singlecell as rsc
 

popv/hub/_metadata.py

@@ -78,8 +78,8 @@ def from_anndata(
             Additional keyword arguments to pass to the HubMetadata initializer.
         """
         setup_dict = adata.uns["_setup_dict"]
-        prediction_keys = adata.uns["prediction_keys"]
-        methods = adata.uns["methods"]
+        prediction_keys = list(adata.uns["prediction_keys"])
+        methods = list(adata.uns["methods"])
         method_kwargs = adata.uns["method_kwargs"]
 
         return cls(
@@ -122,6 +122,8 @@ class HubModelCardHelper:
         The version of `scvi-tools` that the model was trained with.
     anndata_version
         The version of anndata used during model training.
+    popv_version
+        The version of popV that the model was trained with.
     scikit_learn_version
         The version of scikit-learn used during model training.
     organism
@@ -153,6 +155,7 @@ class HubModelCardHelper:
 
     license_info: str
     anndata_version: str
+    popv_version: str
     scikit_learn_version: str
     organism: str
     tissues: list[str] = field(default_factory=list)
@@ -172,6 +175,7 @@ def from_dir(
         license_info: str,
         anndata_version: str,
         scikit_learn_version: str,
+        popv_version: str,
         organism: str,
         metrics_report: str | None = None,
         **kwargs,
@@ -210,6 +214,7 @@ def from_dir(
         return cls(
             license_info,
             anndata_version,
+            popv_version,
             scikit_learn_version,
             organism,
             metrics_report=metrics_report,
@@ -222,11 +227,11 @@ def _to_model_card(self) -> ModelCard:
             "biology",
             "genomics",
             "single-cell",
-            f"anndata_version:{self.anndata_version}",
-            f"scikit_learn_version:{self.scikit_learn_version}",
+            f"AnnData:{self.anndata_version}",
+            f"scikit_learn:{self.scikit_learn_version}",
             f"organism:{self.organism}",
-            f"python_version:{'.'.join([str(i) for i in sys.version_info[:3]])}",
-            "popV",
+            f"Python:{'.'.join([str(i) for i in sys.version_info[:3]])}",
+            f"popV:{self.popv_version}",
         ]
         for t in self.tissues:
             tags.append(f"tissue: {t}")

pyproject.toml

@@ -10,7 +10,7 @@ exclude = ["resources/*", "tests/*", "dataset/*"]
 
 [project]
 name = "popV"
-version = "0.6.0"
+version = "0.6.1"
 description = "Consensus prediction of cell type labels with popV"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -41,16 +41,17 @@ dependencies = [
     "anndata",
     "bbknn",
     "celltypist",
-    "harmonypy",
+    "harmonypy<0.1.0",
     "huggingface-hub",
     "onclass",
     "scanorama",
     "scanpy",
     "scikit-misc",
     "scvi-tools",
-    "tensorflow",
+    "tensorflow<2.18",
     "huggingface",
     "xgboost",
+    "scikit-learn<1.8.0",
     "faiss-cpu",
     "leidenalg",
     "cellxgene_census",

tests/core/test_models.py

@@ -95,7 +95,7 @@ def test_annotation_hub(private: bool):
         "references": "Tabula Sapiens reveals transcription factor expression, senescence effects, and sex-specific features in cell types from 28 human organs and tissues, The Tabula Sapiens Consortium; bioRxiv, doi: https://doi.org/10.1101/2024.12.03.626516",
         "license_info": "cc-by-4.0",
     }
-    hmch = popv.hub.HubModelCardHelper.from_dir(output_folder, anndata_version=anndata.__version__, **model_json)
+    hmch = popv.hub.HubModelCardHelper.from_dir(output_folder, anndata_version=anndata.__version__, popv_version=popv.__version__, **model_json)
     hm = popv.hub.HubMetadata.from_anndata(
         adata,
         popv_version=popv.__version__,