Add tests for S-exon ordering processing

Author: diegozea

tests/test_s_exon_ordering.py

@@ -0,0 +1,87 @@
+import pandas as pd
+import numpy as np
+import pytest
+
+from thoraxe.subexons.tidy import get_tidy_table
+
+
+def _mk_row(gene="G", trx="T", subexon_rank=15, strand=-1, start=0, end=None, label="X"):
+    # End defaults to one base before start on the negative strand or one base after on positive
+    if end is None:
+        end = start - 3 if strand == -1 else start + 3
+
+    # Keep protein length consistent with S_exon_Lengths ("21")
+    subexon_prot = "A" * 21
+    subexon_nt = "ATG" * 7  # 21 aa -> 63 nt; not strictly required but tidy to keep
+
+    # NOTE: get_tidy_table reads S_exon_Sequences (plural) from input rows
+    s_exon_seq = "A" * 21
+
+    return {
+        "GeneID": gene,
+        "TranscriptIDCluster": trx,
+        "SubexonRank": subexon_rank,
+        "Strand": strand,
+        "SubexonCodingStart": int(start),
+        "SubexonCodingEnd": int(end),
+
+        # Columns consumed by tidy.get_tidy_table()
+        "SubexonProteinSequence": subexon_prot,
+        "SubexonSequence": subexon_nt,
+
+        # One s-exon per row in these tests; order is what we care about
+        "S_exons": label,
+        "S_exon_Lengths": "21",
+        "S_exon_Sequences": s_exon_seq,
+    }
+
+
+@pytest.mark.parametrize(
+    "species,trx,strand,starts",
+    [
+        # Platypus (ornithorhynchus_anatinus) transcripts
+        ("ornithorhynchus_anatinus", "ENSOANT00000050045", -1,
+         {"17_2": 29149002, "1_0": 29143989, "1_1": 29143910}),
+        ("ornithorhynchus_anatinus", "ENSOANT00000062336", -1,
+         {"17_2": 29149002, "1_0": 29143989, "1_1": 29143910}),
+        # Xenopus (xenopus_tropicalis) transcript
+        ("xenopus_tropicalis", "ENSXETT00000106354", -1,
+         {"17_2": 3169580, "1_0": 3164877, "1_1": 3164798}),
+    ],
+)
+def test_real_negative_strand_order(species, trx, strand, starts):
+    # Create three rows with the same SubexonRank to trigger tie-breaking
+    # Intentionally insert rows in an order that is NOT genomic order
+    # so the test fails if get_tidy_table does not break ties using
+    # genomic start positions (via the signed-start sort key).
+    insertion_order = ["1_1", "1_0", "17_2"]
+    rows = [
+        _mk_row(
+            gene="G",
+            trx=trx,
+            subexon_rank=15,
+            strand=strand,
+            start=starts[label],
+            end=starts[label] - 1,
+            label=label,
+        )
+        for label in insertion_order
+    ]
+    pre = pd.DataFrame(rows)
+
+    # gene2species mapping is required by get_tidy_table in your ThorAxe branch
+    tidy = get_tidy_table(pre, gene2species={"G": species})
+
+    grp = tidy[tidy["TranscriptIDCluster"] == trx].sort_values("S_exon_Rank")
+    got = grp["S_exonID"].tolist()
+
+    # Expected order: for negative strand, 5′→3′ is decreasing start.
+    # The tidy table exposes `SubexonCodingStart` (not `S_exon_CodingStart`).
+    signed = np.where(
+        grp["Strand"].to_numpy() == 1,
+        grp["SubexonCodingStart"].to_numpy(),
+        -grp["SubexonCodingStart"].to_numpy(),
+    )
+    exp = grp.iloc[np.argsort(signed, kind="stable")]["S_exonID"].tolist()
+
+    assert got == exp, f"{species}:{trx} expected {exp} by genomic order, got {got}"