Merge pull request #131 from CompOmics/ATPs/main

`io.percolator`: Improve parsing of Comet-style peptidoform notation

Author: RalfG

psm_utils/io/percolator.py

@@ -145,9 +145,19 @@ def _infer_charge_columns(fieldnames: list[str]) -> tuple[str | None, dict[int,
     @staticmethod
     def _parse_peptidoform(percolator_peptide: str, charge: int | None) -> Peptidoform:
         """Parse Percolator TSV peptide notation to Peptidoform."""
-        # Remove leading and trailing amino acids
+        # Remove leading and trailing amino acids (e.g., R.PEPTIDE.S -> PEPTIDE)
         match = re.match(r"^(?:[A-Z-])?\.(.+)\.(?:[A-Z-])?$", percolator_peptide)
-        peptidoform = match[1] if match else percolator_peptide
+        peptidoform: str = match[1] if match else percolator_peptide
+
+        # Handle Comet's n-terminal modification format: n[42.0106]PEPTIDE -> [42.0106]-PEPTIDE
+        peptidoform = re.sub(r"^n\[([+-]?[\w\.]*?)\]", r"[\1]-", peptidoform)
+
+        # Handle Comet's c-terminal modification format: PEPTIDEc[-0.9840] -> PEPTIDE-[-0.9840]
+        peptidoform = re.sub(r"c\[([+-]?[\w\.]*?)\]$", r"-[\1]", peptidoform)
+
+        # Ensure positive values inside square brackets have a '+' sign
+        peptidoform = re.sub(r"\[(\d+[\.]*\d*)]", r"[+\1]", peptidoform)
+
         if charge:
             peptidoform += f"/{charge}"
         return Peptidoform(peptidoform)

tests/test_io/test_percolator.py

@@ -28,11 +28,32 @@ def test__infer_charge_columns(self):
 
     def test_parse_peptidoform(self):
         test_cases = [
+            # Basic cases
             (("ACDEFGHR", None), "ACDEFGHR"),
             (("K.ACDEFGHR.I", 1), "ACDEFGHR/1"),
             (("K.ACDEFGHR.-", 2), "ACDEFGHR/2"),
             (("-.ACDEFGHR.I", 3), "ACDEFGHR/3"),
             (("-.ACDEFGHR.-", None), "ACDEFGHR"),
+            # N-terminal modifications
+            (("-.n[42.0106]ACDEFGHR.-", None), "[+42.0106]-ACDEFGHR"),
+            (("n[42.0106]ACDEFGHR", None), "[+42.0106]-ACDEFGHR"),  # Without flanking
+            (("-.n[43]ACDEFGHR.-", 2), "[+43]-ACDEFGHR/2"),  # Integer mass
+            # C-terminal modifications
+            (("-.ACDEFGHRc[-0.9840].-", None), "ACDEFGHR-[-0.984]"),
+            (("ACDEFGHRc[-0.9840]", None), "ACDEFGHR-[-0.984]"),  # Without flanking
+            (("-.ACDEFGHRc[17.0265].-", 2), "ACDEFGHR-[+17.0265]/2"),  # Positive C-term
+            # Internal modifications
+            (("-.ACDEFM[15.9949]GHR.-", None), "ACDEFM[+15.9949]GHR"),
+            (("-.ACDEM[-18.010565]GHR.-", None), "ACDEM[-18.010565]GHR"),  # Negative internal
+            (("-.AC[57.021]DEFGHR.-", None), "AC[+57.021]DEFGHR"),  # Carbamidomethyl
+            # Multiple modifications
+            (("-.n[43]ACDEFM[16]GHR.-", None), "[+43]-ACDEFM[+16]GHR"),  # N-term + internal
+            (("-.ACDEFM[16]GHRc[-1].-", None), "ACDEFM[+16]GHR-[-1]"),  # Internal + C-term
+            (("-.n[42]ACDEFM[16]GHRc[-1].-", 2), "[+42]-ACDEFM[+16]GHR-[-1]/2"),  # All three
+            (("-.AC[57]DEM[16]GHK.-", None), "AC[+57]DEM[+16]GHK"),  # Multiple internal
+            # Already has '+' sign (should not add another)
+            (("-.ACDEFM[+15.9949]GHR.-", None), "ACDEFM[+15.9949]GHR"),
+            (("-.n[+42.0106]ACDEFGHR.-", None), "[+42.0106]-ACDEFGHR"),
         ]
         for test_in, expected_out in test_cases:
             assert expected_out == PercolatorTabReader._parse_peptidoform(*test_in).proforma