From f073e7bf2c0f732a77f35547c8f31f4025e7ee57 Mon Sep 17 00:00:00 2001
From: Sai Nirmayi Yasa <92786623+sainirmayi@users.noreply.github.com>
Date: Wed, 17 Apr 2024 16:39:52 +0530
Subject: [PATCH] FIx dataset simulator (#431)

* explicitly convert non-string objects in .obs slots to categoricals

* Add info about fix in comments

Former-commit-id: 624c44219ad04e350e5eb38ab2beb4b2c9179a03
---
 .../dataset_simulator/script.py                  | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/src/tasks/spatial_decomposition/dataset_simulator/script.py b/src/tasks/spatial_decomposition/dataset_simulator/script.py
index 48daafd374..901d7def5f 100644
--- a/src/tasks/spatial_decomposition/dataset_simulator/script.py
+++ b/src/tasks/spatial_decomposition/dataset_simulator/script.py
@@ -23,9 +23,11 @@
 ## VIASH END
 
 CELLTYPE_MIN_CELLS = 25
+
 # Reading input dataset
 adata = ad.read_h5ad(par['input'])
 
+
 def generate_synthetic_dataset(
     adata: ad.AnnData,
     alpha: Union[float, Sequence] = 1.0,
@@ -162,12 +164,14 @@ def generate_synthetic_dataset(
     adata_merged.uns["cell_type_names"] = uni_labs
     return adata_merged
 
+
 def filter_celltypes(adata, min_cells=CELLTYPE_MIN_CELLS):
     """Filter rare celltypes from an AnnData"""
     celltype_counts = adata.obs["cell_type"].value_counts() >= min_cells
     keep_cells = np.isin(adata.obs["cell_type"], celltype_counts.index[celltype_counts])
     return adata[adata.obs.index[keep_cells]].copy()
 
+
 def filter_genes_cells(adata):
     """Remove empty cells and genes."""
     if "var_names_all" not in adata.uns:
@@ -176,6 +180,7 @@ def filter_genes_cells(adata):
     sc.pp.filter_genes(adata, min_cells=1)
     sc.pp.filter_cells(adata, min_counts=2)
 
+
 adata.X = adata.layers["counts"]
 sc.pp.filter_genes(adata, min_counts=10)
 adata_merged = generate_synthetic_dataset(adata, 
@@ -189,8 +194,15 @@ def filter_genes_cells(adata):
 adata_merged.uns["spatial_data_summary"] = f"Dirichlet alpha={par['alpha']}"
 filter_genes_cells(adata_merged)
 adata_merged.X = None
-if "is_primary_data" in adata_merged.obs:
-    adata_merged.obs['is_primary_data'] = adata_merged.obs['is_primary_data'].fillna(False)
+
+# Convert non-string objects to categoricals to avoid
+# TypeError: Can't implicitly convert non-string objects to strings
+# In this case, the error is raised when there are NA values in .obs columns with dtype object (boolean).
+# The resulting anndata object cannot be written to a file.
+# This conversion is handled in later versions of anndata (0.10)
+for col in adata_merged.obs:
+    if adata_merged.obs[col].dtype == 'object':
+        adata_merged.obs[col] = adata_merged.obs[col].astype('category')
 
 print("Writing output to file")
 adata_merged.write_h5ad(par["simulated_data"])