Updated precommit.

src/cansig/run/integration.py

@@ -10,6 +10,7 @@
 import cansig.filesys as fs  # pytype: disable=import-error
 import cansig.logger as clogger  # pytype: disable=import-error
 import cansig.models.api as models  # pytype: disable=import-error
+from cansig.utils import read_anndata  # pytype: disable=import-error
 
 LOGGER = logging.getLogger(__name__)
 DEFAULT_OUTPUT_BASE_PATH = pathlib.Path("./outputs/batch-integration")
@@ -109,7 +110,7 @@ def integrate(
     fs.save_settings(settings=config, path=output_dir.integration_settings)
 
     # Train the model and get the representations
-    data = anndata.read_h5ad(data_path)
+    data = read_anndata(data_path)
     representations = integrate_adata(data=data, config=config)
 
     # Save the representations

src/cansig/run/metasignatures.py

@@ -6,7 +6,6 @@
 import logging
 
 import anndata as ad  # pytype: disable=import-error
-import scanpy as sc  # pytype: disable=import-error
 import pandas as pd  # pytype: disable=import-error
 import numpy as np  # pytype: disable=import-error
 import pathlib as pl  # pytype: disable=import-error
@@ -24,6 +23,7 @@
 from scanpy.tools._rank_genes_groups import _Method  # pytype: disable=import-error
 
 from cansig.types import Pathlike  # pytype: disable=import-error
+from cansig.utils import read_anndata  # pytype: disable=import-error
 
 _LOGGER = logging.getLogger(__name__)
 _TESTTYPE = Literal["mwu", "ttest"]
@@ -539,7 +539,7 @@ def run_metasignatures(
     _LOGGER.info("Get the information about the signatures in the postprocessing folder.")
     resdict = utils.get_runs_sig(rundir)
 
-    adata = sc.read_h5ad(data_path)
+    adata = read_anndata(data_path)
     # make sure that even in the CanSig case, we don't use the healthy cells
     obs_names = list(pd.concat(resdict["cluster_memb"], axis=1).index)
     adata = adata[obs_names, :].copy()

src/cansig/run/postprocessing.py

@@ -5,7 +5,6 @@
 
 from typing import Union  # pytype: disable=import-error
 
-import anndata  # pytype: disable=import-error
 import pandas as pd  # pytype: disable=import-error
 
 import cansig.cluster.api as cluster  # pytype: disable=import-error
@@ -14,7 +13,7 @@
 import cansig.logger as clogger  # pytype: disable=import-error
 import cansig.plotting.plotting as plotting  # pytype: disable=import-error
 from cansig.gsea import Method  # pytype: disable=import-error
-
+from cansig.utils import read_anndata  # pytype: disable=import-error
 
 LOGGER = logging.getLogger(__name__)
 
@@ -146,7 +145,7 @@ def postprocess(
     # Read the anndata and add the cluster labels
     # TODO(Pawel, Florian, Josephine): Apply preprocessing, e.g., selecting HVGs?
     #  Or maybe this should be in the GEX object?
-    adata = anndata.read_h5ad(data_path)
+    adata = read_anndata(data_path)
     adata = adata[labels.index, :].copy()
     adata.obs[_CLUSTER_COL] = labels
 

src/cansig/utils.py

@@ -0,0 +1,9 @@
+import scanpy as sc  # pytype: disable=import-error
+from cansig.types import Pathlike  # pytype: disable=import-error
+
+
+def read_anndata(path: Pathlike) -> sc.AnnData:
+    # Workaround for a bug in scanpy.
+    adata = sc.read_h5ad(path)
+    adata.uns["log1p"] = {"base": None}
+    return adata