From c7849a76fd0223a576a67827801fa2e20b263d4c Mon Sep 17 00:00:00 2001
From: Kevin Schaper <kevinschaper@gmail.com>
Date: Fri, 5 Jun 2026 13:52:41 -0700
Subject: [PATCH] norm: preserve canonical OBO prefix case in obo_handle
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

`obo_handle` strips `OBO:` from CURIEs like `OBO:FBbt_00000001` and
rebuilds them as `FBbt:00000001`. The final step was uppercasing the
prefix portion with `.upper()`:

    new_id = f"{split_id[0].upper()}:{split_id[1]}"

The accompanying comment ("Capitalize the prefix") describes the intent,
but `.upper()` is the wrong operation: it preserves the case for OBO
namespaces that are conventionally all-uppercase (HP, MONDO, UBERON,
CL, ...) but breaks any mixed-case OBO prefix.

In practice this caused every Drosophila and C. elegans gene-expression
edge in monarch-kg to dangle, because:

  - phenio.json source has canonical IRIs like
    http://purl.obolibrary.org/obo/FBbt_NNNNNNNN.
  - KGX (which lacks an explicit prefix entry for FBbt / WBbt in its
    default JSON-LD context) falls back to writing them as
    `OBO:FBbt_NNNNNNNN`.
  - This function then `.upper()`-ed the prefix, producing `FBBT:` and
    `WBBT:` — neither of which is the canonical OBO Foundry CURIE for
    those namespaces (`FBbt:`, `WBbt:`).
  - Downstream consumers like the Alliance gene-to-expression ingest
    look up nodes by the canonical form, so all ~503k Drosophila +
    C. elegans gene-expression edges per monarch-kg release went to
    the dangling-edge pile.

Drop the `.upper()` and preserve the case as it appears in the OBO
PURL local-name segment. Behaviour is unchanged for all-uppercase OBO
prefixes (HP/MONDO/UBERON/CL/...) — the `.upper()` was a no-op for those
already — and now correct for mixed-case prefixes (FBbt/WBbt/ZFA/XAO/...).

Add regression tests asserting both behaviours.

Verified end-to-end on a synthetic input through `clean_and_normalize_graph`:

  before: OBO:FBbt_00000001 → FBBT:00000001
  after:  OBO:FBbt_00000001 → FBbt:00000001
  HP / MONDO / UBERON / CL unchanged.
---
 tests/test_norm.py    | 30 ++++++++++++++++++++++++++++++
 universalizer/norm.py | 13 +++++++++----
 2 files changed, 39 insertions(+), 4 deletions(-)

diff --git a/tests/test_norm.py b/tests/test_norm.py
index bb83fb0..11ae4ed 100644
--- a/tests/test_norm.py
+++ b/tests/test_norm.py
@@ -86,3 +86,33 @@ def test_normalize_and_remove_extra_edges(self):
         with open(self.chebi_graph_edges, "r") as infile:
             afterlen = len(infile.readlines())
         self.assertLess(afterlen, beforelen, "Edges not removed?")
+
+
+class TestObohandle(TestCase):
+    """Tests for `obo_handle`'s prefix-casing behaviour."""
+
+    def test_uppercase_prefix_unchanged(self):
+        """Pure-uppercase OBO prefixes survive obo_handle as-is.
+
+        HP, MONDO, UBERON, CL, etc. were already uppercase, so the previous
+        `.upper()` was a no-op for them. Removing that call must not regress
+        the no-op.
+        """
+        from universalizer.norm import obo_handle
+        self.assertEqual(obo_handle("OBO:HP_0000118"), "HP:0000118")
+        self.assertEqual(obo_handle("OBO:MONDO_0000001"), "MONDO:0000001")
+        self.assertEqual(obo_handle("OBO:UBERON_0000061"), "UBERON:0000061")
+        self.assertEqual(obo_handle("OBO:CL_0000000"), "CL:0000000")
+
+    def test_mixed_case_prefix_preserved(self):
+        """Mixed-case OBO prefixes (FBbt, WBbt) keep their canonical case.
+
+        Regression test for the bug where `obo_handle` `.upper()`-ed the
+        prefix, turning `OBO:FBbt_00000001` into `FBBT:00000001` instead of
+        the canonical `FBbt:00000001`. This silently dropped every
+        Drosophila and C. elegans gene-expression edge in downstream
+        consumers (e.g. monarch-kg's alliance_gene_to_expression).
+        """
+        from universalizer.norm import obo_handle
+        self.assertEqual(obo_handle("OBO:FBbt_00000001"), "FBbt:00000001")
+        self.assertEqual(obo_handle("OBO:WBbt_0000001"), "WBbt:0000001")
diff --git a/universalizer/norm.py b/universalizer/norm.py
index 574ca3f..02ca6af 100644
--- a/universalizer/norm.py
+++ b/universalizer/norm.py
@@ -502,10 +502,15 @@ def obo_handle(old_id: str) -> str:
         # Replace the first underscore with colon
         new_id = new_id.replace("_", ":", 1)
 
-        # Capitalize the prefix, but not the rest
-        # of the ID
-        split_id = new_id.split(":", 1)
-        new_id = f"{split_id[0].upper()}:{split_id[1]}"
+        # Preserve the canonical case of the prefix as it appears in the
+        # source OBO PURL — `FBbt`, `WBbt` etc. stay mixed-case; pure
+        # uppercase prefixes (HP, MONDO, UBERON, CL, ...) are unchanged.
+        # Previously this `.upper()`-ed the whole prefix, which broke any
+        # mixed-case OBO namespace by emitting e.g. `FBBT:` instead of the
+        # canonical `FBbt:`. That caused every Drosophila and C. elegans
+        # gene-expression edge in monarch-kg to dangle (~503k edges per
+        # release on `alliance_gene_to_expression_edges`), because Alliance
+        # was looking up nodes by the canonical mixed-case CURIE.
     else:
         new_id = old_id