pybiolib 1.2.158.dev1__py3-none-any.whl → 1.2.163.dev1__py3-none-any.whl

biolib/utils/seq_util.py

@@ -1,7 +1,7 @@
 import re
 from io import BufferedIOBase, TextIOBase
 
-from biolib.typing_utils import Dict, List, Optional, Union, Iterator
+from biolib.typing_utils import Dict, Iterator, List, Optional, Union
 
 
 class SeqUtilRecord:
@@ -35,14 +35,22 @@ class SeqUtil:
         input_file: Union[str, BufferedIOBase, None] = None,
         default_header: Optional[str] = None,
         allow_any_sequence_characters: bool = False,
+        use_strict_alphabet: Optional[bool] = False,
         allow_empty_sequence: bool = True,
         file_name: Optional[str] = None,
 ) -> Iterator[SeqUtilRecord]:
         def process_and_yield_record(header: str, sequence_lines: List[str]):
             sequence = ''.join(sequence_lines)
             sequence_id = header.split()[0]
+            if allow_any_sequence_characters and use_strict_alphabet:
+                raise Exception(
+                    'Error: Please choose either allow_any_sequence_characters or use_strict_alphabet'
+                )
             if not allow_any_sequence_characters:
-                invalid_sequence_characters = SeqUtil._find_invalid_sequence_characters(sequence)
+                if use_strict_alphabet:
+                    invalid_sequence_characters = SeqUtil._find_invalid_sequence_characters_strict(sequence)
+                else:
+                    invalid_sequence_characters = SeqUtil._find_invalid_sequence_characters(sequence)
                 if invalid_sequence_characters:
                     raise Exception(
                         f'Error: Invalid character ("{invalid_sequence_characters[0]}") found in sequence {sequence_id}'
@@ -126,6 +134,15 @@ class SeqUtil:
         invalid_chars = [char for char in sequence if char not in allowed_sequence_chars]
         return invalid_chars
 
+    @staticmethod
+    def _find_invalid_sequence_characters_strict(sequence: str) -> List[str]:
+        # Equivalent to fair-esm alphabet, compatible with ESM-models
+        # Excludes digits, '_' and 'J' (ambiguous letter only used in mass-spec NMR)
+        # https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/esm/constants.py#L8
+        allowed_sequence_chars = set('lagvsertidpkqnfymhwcxbuzoLAGVSERTIDPKQNFYMHWCXBUZO-.')
+        invalid_chars = [char for char in sequence if char not in allowed_sequence_chars]
+        return invalid_chars
+
     @staticmethod
     def _find_invalid_sequence_id_characters(sequence: str) -> List[str]:
         allowed_chars = set('abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-_.:*#')

{pybiolib-1.2.158.dev1.dist-info → pybiolib-1.2.163.dev1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pybiolib
-Version: 1.2.158.dev1
+Version: 1.2.163.dev1
 Summary: BioLib Python Client
 Home-page: https://github.com/biolib
 License: MIT

{pybiolib-1.2.158.dev1.dist-info → pybiolib-1.2.163.dev1.dist-info}/RECORD

@@ -117,10 +117,10 @@ biolib/utils/__init__.py,sha256=fwjciJyJicvYyZcVTzfDBgD0SKY13DeXqvTeG4qZIy8,5548
 biolib/utils/app_uri.py,sha256=Yq_-_VGugQhMMo6mM5f0G9yNlLkr0WK4j0Nrf3FE4xQ,2171
 biolib/utils/cache_state.py,sha256=u256F37QSRIVwqKlbnCyzAX4EMI-kl6Dwu6qwj-Qmag,3100
 biolib/utils/multipart_uploader.py,sha256=XvGP1I8tQuKhAH-QugPRoEsCi9qvbRk-DVBs5PNwwJo,8452
-biolib/utils/seq_util.py,sha256=WJnU9vZdwY8RHXvzATyV80OXzyJ7w9EkG33Tna9Nr6A,5698
+biolib/utils/seq_util.py,sha256=Ozk0blGtPur_D9MwShD02r_mphyQmgZkx-lOHOwnlIM,6730
 biolib/utils/zip/remote_zip.py,sha256=0wErYlxir5921agfFeV1xVjf29l9VNgGQvNlWOlj2Yc,23232
-pybiolib-1.2.158.dev1.dist-info/LICENSE,sha256=F2h7gf8i0agDIeWoBPXDMYScvQOz02pAWkKhTGOHaaw,1067
-pybiolib-1.2.158.dev1.dist-info/METADATA,sha256=7gYFVnp6YM8fLfHzg0S_XDeC6DxdTzeDX6dfWesenm0,1512
-pybiolib-1.2.158.dev1.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-pybiolib-1.2.158.dev1.dist-info/entry_points.txt,sha256=p6DyaP_2kctxegTX23WBznnrDi4mz6gx04O5uKtRDXg,42
-pybiolib-1.2.158.dev1.dist-info/RECORD,,
+pybiolib-1.2.163.dev1.dist-info/LICENSE,sha256=F2h7gf8i0agDIeWoBPXDMYScvQOz02pAWkKhTGOHaaw,1067
+pybiolib-1.2.163.dev1.dist-info/METADATA,sha256=JK1zdgar9NsD7Wv30JWkX3NNIzcKLQHQVT-Hy6n-I38,1512
+pybiolib-1.2.163.dev1.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+pybiolib-1.2.163.dev1.dist-info/entry_points.txt,sha256=p6DyaP_2kctxegTX23WBznnrDi4mz6gx04O5uKtRDXg,42
+pybiolib-1.2.163.dev1.dist-info/RECORD,,