Add allen brain

src/data_processors/process_dataset/script.py

@@ -213,6 +213,7 @@ def subsample_adata_group_balanced(adata, group_key, n_samples, seed=0):
 if "metadata" not in sdata.tables:
     if "table" in sdata.tables:
         sdata["metadata"] = sdata["table"]
+        del sdata["table"]
     else:
         sdata["metadata"] = ad.AnnData(uns={})
 

src/datasets/loaders/allen_brain_cell_atlas_merfish/config.vsh.yaml

@@ -91,4 +91,4 @@ runners:
   - type: executable
   - type: nextflow
     directives:
-      label: [midmem, midcpu, hightime]
+      label: [highmem, midcpu, hightime]

src/datasets/loaders/allen_brain_cell_atlas_merfish/script.py

@@ -344,7 +344,9 @@ def um_to_px_y(y):
         fov_img = fov_img[:, OVERLAP_SIZE:, :]
         xp += OVERLAP_SIZE
 
-    complete_img[:, xp:xp + fov_img.shape[1], yp:yp + fov_img.shape[2]] = fov_img
+    h = min(fov_img.shape[1], complete_img.shape[1] - xp)
+    w = min(fov_img.shape[2], complete_img.shape[2] - yp)
+    complete_img[:, xp:xp + h, yp:yp + w] = fov_img[:, :h, :w]
 
     if fov % 10 == 0:
         delete_dax_files()

src/datasets/loaders/bruker_cosmx/config.vsh.yaml

@@ -76,4 +76,4 @@ runners:
   - type: executable
   - type: nextflow
     directives:
-      label: [midmem, midcpu, hightime]
+      label: [highmem, midcpu, hightime]

src/methods_segmentation/binning/script.py

@@ -49,8 +49,8 @@ def convert_to_lower_dtype(arr):
 parsed_data = Labels2DModel.parse(data_array, transformations=transformation)
 sd_output.labels['segmentation'] = parsed_data
 sd_output.tables['table'] = ad.AnnData(
-      obs=sdata.tables["table"].obs[["cell_id", "region"]],
-      var=sdata.tables["table"].var[[]]
+      obs=sdata.tables["metadata"].obs[["cell_id", "region"]],
+      var=sdata.tables["metadata"].var[[]]
     )
 
 print("Writing output", flush=True)

src/methods_segmentation/cellpose/config.vsh.yaml

@@ -79,8 +79,8 @@ resources:
 
 engines:
   - type: docker
-    #image: openproblems/base_pytorch_nvidia:1 # TODO: ideally get gpu image to work
-    image: openproblems/base_python:1
+    image: openproblems/base_pytorch_nvidia:1 # TODO: ideally get gpu image to work
+    #image: openproblems/base_python:1
     setup:
       - type: python
         pypi: cellpose<4.0.0
@@ -93,4 +93,4 @@ runners:
   - type: executable
   - type: nextflow
     directives:
-      label: [ midtime, midcpu, veryhighmem, gpu ]
+      label: [ midtime, midcpu, highmem, gpu ]

src/methods_segmentation/cellpose/script.py

@@ -48,8 +48,8 @@ def convert_to_lower_dtype(arr):
 sd_output.labels['segmentation'] = parsed_data
 
 sd_output.tables['table'] = ad.AnnData(
-      obs=sdata.tables["table"].obs[["cell_id", "region"]],
-      var=sdata.tables["table"].var[[]]
+      obs=sdata.tables["metadata"].obs[["cell_id", "region"]],
+      var=sdata.tables["metadata"].var[[]]
     )
 
 print("Writing output", flush=True)

src/methods_segmentation/custom_segmentation/script.py

@@ -26,8 +26,8 @@
   },
   tables={
     "table": ad.AnnData(
-      obs=sdata.tables["table"].obs[["cell_id", "region"]],
-      var=sdata.tables["table"].var[[]]
+      obs=sdata.tables["metadata"].obs[["cell_id", "region"]],
+      var=sdata.tables["metadata"].var[[]]
     )
   }
 )

src/methods_segmentation/stardist/config.vsh.yaml

@@ -24,21 +24,20 @@ engines:
   # NOTE: On mac the tensorflow install leads to an error. Develop the method in a conda env instead (and test docker via gh-actions). 
   #       Installations can be done with pip (except tensorflow: use conda install -c conda-forge tensorflow)
   - type: docker
-    #image: openproblems/base_tensorflow_nvidia:1 # TODO: ideally get gpu image to work
-    image: openproblems/base_python:1
+    image: openproblems/base_tensorflow_nvidia:1 # TODO: ideally get gpu image to work
+    #image: openproblems/base_python:1
     __merge__: 
       - /src/base/setup_spatialdata_partial.yaml
     setup:
       - type: python
-        pypi: stardist
-      - type: python
-        pypi: numpy<2.0.0
-      - type: python
-        pypi: tensorflow
+        pypi:
+          - stardist
+          - tensorflow==2.17.0
+          - numpy<2.0.0
   - type: native
 
 runners:
   - type: executable
   - type: nextflow
     directives:
-      label: [ midtime, midcpu, highmem, gpu ]
+      label: [ hightime, midcpu, highmem, gpu ]

src/methods_segmentation/stardist/script.py

@@ -78,8 +78,8 @@ def do_after(self):
 sd_output.labels['segmentation'] = parsed_labels
 
 sd_output.tables['table'] = ad.AnnData(
-      obs=sdata.tables["table"].obs[["cell_id", "region"]],
-      var=sdata.tables["table"].var[[]]
+      obs=sdata.tables["metadata"].obs[["cell_id", "region"]],
+      var=sdata.tables["metadata"].var[[]]
     )
 
 print("Writing output", flush=True)

src/methods_segmentation/watershed/script.py

@@ -49,8 +49,8 @@ def convert_to_lower_dtype(arr):
 sd_output.labels['segmentation'] = parsed_data
 
 sd_output.tables['table'] = ad.AnnData(
-      obs=sdata.tables["table"].obs[["cell_id", "region"]],
-      var=sdata.tables["table"].var[[]]
+      obs=sdata.tables["metadata"].obs[["cell_id", "region"]],
+      var=sdata.tables["metadata"].var[[]]
     )
 
 print("Writing output", flush=True)

src/methods_transcript_assignment/baysor/script_no_sopa.py

@@ -177,7 +177,7 @@ def convert_str_ids_to_ints(df, file_path_for_error_messages=None):
     tables={
         "table": ad.AnnData(
           obs=pd.DataFrame(cell_id_col),
-          var=sdata.tables["table"].var[[]]
+          var=sdata.tables["metadata"].var[[]]
         )
     }
 )

src/methods_transcript_assignment/proseg/script_no_sopa.py

@@ -132,7 +132,7 @@
     tables={
         "table": ad.AnnData(
           obs=pd.DataFrame(cell_id_col),
-          var=sdata.tables["table"].var[[]]
+          var=sdata.tables["metadata"].var[[]]
         )
     }
 )