Fix Celltyping OOM (#156)

nextflow/modules/nf-core/celltyping/main.nf

@@ -1,6 +1,6 @@
 process CELLTYPING {
     tag "$meta.id"
-    label 'process_low'
+    label 'process_high_memory'
     container "${ workflow.containerEngine == 'apptainer' && !task.ext.singularity_pull_docker_container ?
         ('docker://jeffquinnmsk/nuc2seg:' + params.nuc2seg_version) :
         ('docker.io/jeffquinnmsk/nuc2seg:' + params.nuc2seg_version) }"

src/nuc2seg/celltyping.py

@@ -50,7 +50,7 @@ def fit_celltype_em_model(
     gene_counts,
     gene_names,
     min_components=2,
-    max_components=25,
+    max_components=20,
     max_em_steps=100,
     tol=1e-4,
     warm_start=False,
@@ -270,7 +270,7 @@ def fit_celltyping_on_segments_and_transcripts(
     tx_geo_df: geopandas.GeoDataFrame,
     foreground_nucleus_distance: float = 1,
     min_components: int = 2,
-    max_components: int = 25,
+    max_components: int = 20,
     rng: np.random.Generator = None,
 ):
     # Create a nuclei x gene count matrix

src/nuc2seg/celltyping_test.py

@@ -67,7 +67,7 @@ def test_estimate_cell_types2():
         data,
         gene_names=[f"gene_{i}" for i in range(n_genes)],
         min_components=2,
-        max_components=25,
+        max_components=5,
         max_em_steps=10,
         tol=1e-4,
         warm_start=False,
@@ -88,8 +88,8 @@ def test_estimate_cell_types2():
         celltyping_results.relative_expression,
     )
 
-    assert len(aic_scores) == 24
-    assert len(bic_scores) == 24
+    assert len(aic_scores) == 4
+    assert len(bic_scores) == 4
     assert aic_scores.argmin() == bic_scores.argmin() == 1
 
 

src/nuc2seg/cli/celltyping.py

@@ -77,7 +77,7 @@ def get_parser():
         "--max-n-celltypes",
         help="Maximum number of cell types to consider (inclusive).",
         type=int,
-        default=25,
+        default=20,
     )
     parser.add_argument(
         "--min-n-celltypes",