Update emd vertical when it cannot be calculated

Author: ghar1821

src/metrics/emd/helper.py

@@ -50,35 +50,48 @@ def calculate_vertical_emd(
         i_adata=i_split2_adata, markers_to_assess=markers_to_assess
     )
 
-    emd_long = pd.concat([emd_split1_long, emd_split2_long])
-
-    # mean global emd across all sample combinations, markers, and splits
-    mean_emd_global = np.nanmean(
-        emd_long[emd_long["cell_type"] == "global"]
-        .drop(columns=["cell_type", "first_sample", "second_sample"])
-        .to_numpy()
-        .flatten()
-    )
-    max_emd_global = np.nanmax(
-        emd_long[emd_long["cell_type"] == "global"]
-        .drop(columns=["cell_type", "first_sample", "second_sample"])
-        .to_numpy()
-        .flatten()
-    )
+    # safeguard
+    mean_emd_global = np.nan
+    max_emd_global = np.nan
+    mean_emd_ct = np.nan
+    max_emd_ct = np.nan
+
+    # compute these only if we can.
+    emd_long = []
+    for df in [emd_split1_long, emd_split2_long]:
+        if isinstance(df, pd.DataFrame):
+            emd_long.append(df)
+
+    if len(emd_long) > 0:
+        emd_long = pd.concat(emd_long)
+
+        # mean global emd across all sample combinations, markers, and splits
+        mean_emd_global = np.nanmean(
+            emd_long[emd_long["cell_type"] == "global"]
+            .drop(columns=["cell_type", "first_sample", "second_sample"])
+            .to_numpy()
+            .flatten()
+        )
+        max_emd_global = np.nanmax(
+            emd_long[emd_long["cell_type"] == "global"]
+            .drop(columns=["cell_type", "first_sample", "second_sample"])
+            .to_numpy()
+            .flatten()
+        )
 
-    # mean cell type emd across all sample combinations, markers, and splits
-    mean_emd_ct = np.nanmean(
-        emd_long[emd_long["cell_type"] != "global"]
-        .drop(columns=["cell_type", "first_sample", "second_sample"])
-        .to_numpy()
-        .flatten()
-    )
-    max_emd_ct = np.nanmax(
-        emd_long[emd_long["cell_type"] != "global"]
-        .drop(columns=["cell_type", "first_sample", "second_sample"])
-        .to_numpy()
-        .flatten()
-    )
+        # mean cell type emd across all sample combinations, markers, and splits
+        mean_emd_ct = np.nanmean(
+            emd_long[emd_long["cell_type"] != "global"]
+            .drop(columns=["cell_type", "first_sample", "second_sample"])
+            .to_numpy()
+            .flatten()
+        )
+        max_emd_ct = np.nanmax(
+            emd_long[emd_long["cell_type"] != "global"]
+            .drop(columns=["cell_type", "first_sample", "second_sample"])
+            .to_numpy()
+            .flatten()
+        )
 
     return {
         KEY_MEAN_EMD_GLOBAL: mean_emd_global,
@@ -122,10 +135,10 @@ def get_vert_emd_for_integrated_adata(i_adata: ad.AnnData, markers_to_assess: li
 
         print(
             f"{i_adata.uns['dataset_id']} from {i_adata.uns['method_id']} does not have"
-            f"at least 2 samples per group. Skipping EMD vertical calculation."
+            f" at least 2 samples per group. Skipping EMD vertical calculation."
         )
 
-        return np.nan, np.nan, np.nan
+        return np.nan, np.nan
 
     cell_types = i_adata.obs["cell_type"].unique()
 

src/metrics/emd/script.py

@@ -9,16 +9,21 @@
     "input_integrated_split1": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split1.h5ad",
     "input_integrated_split2": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split2.h5ad",
     "input_unintegrated": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/unintegrated.h5ad",
+    # "input_integrated_split1": "resources_test/task_cyto_batch_integration/cytonorm_data_full/cycombine_mid_out.h5ad",
+    # "input_integrated_split2": "resources_test/task_cyto_batch_integration/cytonorm_data_full/cycombine_mid_out.h5ad",
+    # "input_unintegrated": "resources_test/task_cyto_batch_integration/cytonorm_data_full/unintegrated.h5ad",
     "output": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/emd_out.h5ad",
 }
-meta = {"name": "emd", "resources_dir": "src/utils/helper_functions.py"}
+meta = {"name": "emd", "resources_dir": "src/utils/"}
 ## VIASH END
 
 sys.path.append(meta["resources_dir"])
 
 import helper as emd_helper
 import helper_functions as global_helper
 
+# import src.metrics.emd.helper as emd_helper
+
 print("Reading input files", flush=True)
 
 input_integrated_split1 = ad.read_h5ad(par["input_integrated_split1"])
@@ -37,11 +42,15 @@
 )
 
 # more preprocessing
-input_integrated_split1 = global_helper.subset_markers_tocorrect(input_integrated_split1)
+input_integrated_split1 = global_helper.subset_markers_tocorrect(
+    input_integrated_split1
+)
 input_integrated_split1 = global_helper.subset_nocontrols(input_integrated_split1)
 input_integrated_split1 = global_helper.remove_unlabelled(input_integrated_split1)
 
-input_integrated_split2 = global_helper.subset_markers_tocorrect(input_integrated_split2)
+input_integrated_split2 = global_helper.subset_markers_tocorrect(
+    input_integrated_split2
+)
 input_integrated_split2 = global_helper.subset_nocontrols(input_integrated_split2)
 input_integrated_split2 = global_helper.remove_unlabelled(input_integrated_split2)
 
@@ -63,9 +72,9 @@
 # check that the data for each donor in integrated left and right are actually from two different batches!
 for donor in donor_list:
     # donor = donor_list[0]
-    batch_split1 = input_integrated_split1.obs[input_integrated_split1.obs["donor"] == donor][
-        "batch"
-    ].unique()
+    batch_split1 = input_integrated_split1.obs[
+        input_integrated_split1.obs["donor"] == donor
+    ]["batch"].unique()
     batch_split2 = input_integrated_split2.obs[
         input_integrated_split2.obs["donor"] == donor
     ]["batch"].unique()