Fix inverted span for minus-strand multi-exon GenBank locations

src/datasets/packaged_modules/genbank/genbank.py

@@ -310,10 +310,17 @@ def _parse_location_node(cls, text: str) -> dict:
             located = [part for part in parts if "start" in part]
             if located:
                 location["parts"] = [[part["start"], part["end"]] for part in located]
-                location["start"] = located[0]["start"]
-                location["end"] = located[-1]["end"]
-                location["start_partial"] = located[0]["start_partial"]
-                location["end_partial"] = located[-1]["end_partial"]
+                # Parts are listed in the file's (transcription) order, which is descending
+                # genomic order for minus-strand features (NCBI/Biopython write a two-exon
+                # minus-strand CDS as ``complement(join(<high>..<higher>,<low>..<lower>))``).
+                # The aggregate span is therefore min-start/max-end over all parts, matching
+                # Biopython's CompoundLocation.start/.end, not the first/last part positionally.
+                start_part = min(located, key=lambda part: part["start"])
+                end_part = max(located, key=lambda part: part["end"])
+                location["start"] = start_part["start"]
+                location["end"] = end_part["end"]
+                location["start_partial"] = start_part["start_partial"]
+                location["end_partial"] = end_part["end_partial"]
             return location
 
         location = {"strand": 1}

tests/packaged_modules/test_genbank.py

@@ -1098,3 +1098,18 @@ def test_genbank_features_subset_selects_columns(tmp_path):
     features = Features({"locus_name": Value("string"), "sequence": Value("large_string")})
     table = next(iter(GenBank(features=features)._generate_tables([[str(filename)]])))[1]
     assert table.column_names == ["locus_name", "sequence"]
+
+
+def test_genbank_minus_strand_multi_exon_join_span_is_min_max(tmp_path):
+    """A minus-strand multi-exon feature is written by NCBI/Biopython as
+    ``complement(join(<high>..<higher>,<low>..<lower>))`` (exons in transcription order,
+    i.e. descending genomic order). The aggregate span must still be min-start/max-end,
+    as Biopython's CompoundLocation reports, not first-part-start/last-part-end."""
+    _, features = _parse_one(tmp_path, _HDR + "     CDS             complement(join(121..180,11..60))\n" + _ORIGIN)
+    location = features[0]["location"]
+    assert location["strand"] == -1
+    assert location["operator"] == "join"
+    assert location["parts"] == [[121, 180], [11, 60]]
+    assert location["start"] == 11
+    assert location["end"] == 180
+    assert location["start"] <= location["end"]