remove threshold setting as it isn't relevant for tangram projection type used

Author: ctastad

cholab-prep/example_xenium-multichannel-config.yaml

@@ -81,7 +81,6 @@ annotation:
     sc_reference_path: /sc/arion/projects/untreatedIBD/cache/reference-objects/oliver-2024-subset.h5ad # path to the annotated scRNAseq reference used by Tangram
     cell_type_key: level_2_annot # key of adata_ref.obs containing the cell-type annotation
     reference_preprocessing: log1p # if your reference `adata.X` contains raw counts, then remove this parameter from the config. Else, choose "log1p" (if you have already run both `sc.pp.normalize_total` and `sc.pp.log1p` on your reference) or "normalized" (if you have already run `sc.pp.normalize_total` on your reference)
-    threshold: 0.5  # Add this parameter to control the threshold for cell type assignment
 
   # [only if you choose the simple staining-based annotation (comment out the lines below)]
   # method: fluorescence

sopa/utils/annotation.py

@@ -59,7 +59,6 @@ def tangram_annotate(
     reference_preprocessing: str | None = None,
     bag_size: int = 10_000,
     max_obs_reference: int = 10_000,
-    threshold: float = 0.5,
     **kwargs,
 ):
     """Tangram multi-level annotation. Tangram is run on multiple bags of cells to decrease the RAM usage.
@@ -83,7 +82,6 @@ def tangram_annotate(
         reference_preprocessing,
         bag_size,
         max_obs_reference,
-        threshold=threshold,
         **kwargs,
     ).run()
 
@@ -97,7 +95,6 @@ def __init__(
         reference_preprocessing: str | None,
         bag_size: int,
         max_obs_reference: int,
-        threshold: float = 0.5,
         clip_percentile: float = 0.95,
     ):
         self.ad_sp = ad_sp
@@ -112,7 +109,6 @@ def __init__(
         self.reference_preprocessing = reference_preprocessing
         self.bag_size = bag_size
         self.max_obs_reference = max_obs_reference
-        self.threshold = threshold
         self.clip_percentile = clip_percentile
 
         assert (
@@ -284,9 +280,7 @@ def run_group(self, level: int = 0, indices_sp=None, indices_sc=None):
                 device=self.device,
             )
 
-            tg.project_cell_annotations(
-                ad_map, ad_sp_split, annotation=self.level_obs_key(level), threshold=self.threshold
-            )
+            tg.project_cell_annotations(ad_map, ad_sp_split, annotation=self.level_obs_key(level))
 
             res = ad_sp_split.obsm["tangram_ct_pred"]
             self.ad_sp.obsm[self.probs_key(level)].loc[split, res.columns] = res