diff --git a/kg_phenio/normalize.py b/kg_phenio/normalize.py
index 71cf5dc..fdf26a4 100644
--- a/kg_phenio/normalize.py
+++ b/kg_phenio/normalize.py
@@ -12,13 +12,23 @@ def normalize() -> None:
         None
 
     """
+    # Use the Monarch-specific `merged.monarch` context from prefixmaps as
+    # the single source of CURIE↔IRI conventions. The previous
+    # `["obo", "bioregistry.upper"]` pair interacted with universalizer's
+    # lowercased-reverse-IRI logic in `make_id_maps` and was producing
+    # all-uppercase prefix casings for OBO mixed-case anatomies (FBbt →
+    # FBBT, WBbt → WBBT). That dropped ~503k Drosophila + C. elegans
+    # gene-expression edges to dangling in the 2026-06-03 monarch-kg build
+    # (the entire shortfall on `alliance_gene_to_expression_edges`).
+    # `merged.monarch` is also what monarch-app's `curie_service.py` uses,
+    # so the whole stack is now aligned.
     print("Normalizing nodes and categories...")
     clean_and_normalize_graph(
         filepath="data/merged/",
         compressed=False,
         maps=[],
         update_categories=True,
-        contexts=["obo", "bioregistry.upper"],
+        contexts=["merged.monarch"],
         namespace_cat_map="",
         oak_lookup=False,
     )