cool-seq-tool 0.4.0.dev3__py3-none-any.whl → 0.5.0__py3-none-any.whl

cool_seq_tool/sources/transcript_mappings.py

@@ -1,15 +1,15 @@
 """Provide mappings between gene symbols and RefSeq + Ensembl transcript accessions."""
+
 import csv
 from pathlib import Path
-from typing import Dict, List, Optional
 
-from cool_seq_tool.paths import LRG_REFSEQGENE_PATH, TRANSCRIPT_MAPPINGS_PATH
+from cool_seq_tool.resources.data_files import DataFile, get_data_file
 
 
 class TranscriptMappings:
     """Provide mappings between gene symbols and RefSeq + Ensembl transcript accessions.
 
-    Uses ``LRG_RefSeqGene`` and ``transcript_mappings.csv``, which will automatically
+    Uses ``LRG_RefSeqGene`` and ``transcript_mappings.tsv``, which will automatically
     be acquired if they aren't already available. See the
     :ref:`configuration <configuration>` section in the documentation for information
     about manual acquisition of data.
@@ -21,44 +21,53 @@ class TranscriptMappings:
 
     def __init__(
         self,
-        transcript_file_path: Path = TRANSCRIPT_MAPPINGS_PATH,
-        lrg_refseqgene_path: Path = LRG_REFSEQGENE_PATH,
+        transcript_file_path: Path | None = None,
+        lrg_refseqgene_path: Path | None = None,
+        from_local: bool = False,
     ) -> None:
         """Initialize the transcript mappings class.
 
         :param transcript_file_path: Path to transcript mappings file
         :param lrg_refseqgene_path: Path to LRG RefSeqGene file
+        :param from_local: if ``True``, don't check for or acquire latest version --
+            just provide most recent locally available file, if possible, and raise
+            error otherwise
         """
         # ENSP <-> Gene Symbol
-        self.ensembl_protein_version_for_gene_symbol: Dict[str, List[str]] = {}
-        self.ensembl_protein_version_to_gene_symbol: Dict[str, str] = {}
-        self.ensembl_protein_for_gene_symbol: Dict[str, List[str]] = {}
-        self.ensembl_protein_to_gene_symbol: Dict[str, str] = {}
+        self.ensembl_protein_version_for_gene_symbol: dict[str, list[str]] = {}
+        self.ensembl_protein_version_to_gene_symbol: dict[str, str] = {}
+        self.ensembl_protein_for_gene_symbol: dict[str, list[str]] = {}
+        self.ensembl_protein_to_gene_symbol: dict[str, str] = {}
 
         # Gene Symbol <-> ENST
-        self.ensembl_transcript_version_for_gene_symbol: Dict[str, List[str]] = {}
-        self.ensembl_transcript_version_to_gene_symbol: Dict[str, str] = {}
-        self.ensembl_transcript_for_gene_symbol: Dict[str, List[str]] = {}
-        self.ensembl_transcript_to_gene_symbol: Dict[str, str] = {}
+        self.ensembl_transcript_version_for_gene_symbol: dict[str, list[str]] = {}
+        self.ensembl_transcript_version_to_gene_symbol: dict[str, str] = {}
+        self.ensembl_transcript_for_gene_symbol: dict[str, list[str]] = {}
+        self.ensembl_transcript_to_gene_symbol: dict[str, str] = {}
 
         # NP_ <-> Gene Symbol
-        self.refseq_protein_for_gene_symbol: Dict[str, List[str]] = {}
-        self.refseq_protein_to_gene_symbol: Dict[str, str] = {}
+        self.refseq_protein_for_gene_symbol: dict[str, list[str]] = {}
+        self.refseq_protein_to_gene_symbol: dict[str, str] = {}
 
         # NM_ <-> Gene Symbol
-        self.refseq_rna_version_for_gene_symbol: Dict[str, List[str]] = {}
-        self.refseq_rna_version_to_gene_symbol: Dict[str, str] = {}
-        self.refseq_rna_for_gene_symbol: Dict[str, List[str]] = {}
-        self.refseq_rna_to_gene_symbol: Dict[str, str] = {}
+        self.refseq_rna_version_for_gene_symbol: dict[str, list[str]] = {}
+        self.refseq_rna_version_to_gene_symbol: dict[str, str] = {}
+        self.refseq_rna_for_gene_symbol: dict[str, list[str]] = {}
+        self.refseq_rna_to_gene_symbol: dict[str, str] = {}
 
         # NP -> NM
-        self.np_to_nm: Dict[str, str] = {}
+        self.np_to_nm: dict[str, str] = {}
 
         # ENSP -> ENST
-        self.ensp_to_enst: Dict[str, str] = {}
+        self.ensp_to_enst: dict[str, str] = {}
 
-        self._load_transcript_mappings_data(transcript_file_path)
-        self._load_refseq_gene_symbol_data(lrg_refseqgene_path)
+        self._load_transcript_mappings_data(
+            transcript_file_path
+            or get_data_file(DataFile.TRANSCRIPT_MAPPINGS, from_local)
+        )
+        self._load_refseq_gene_symbol_data(
+            lrg_refseqgene_path or get_data_file(DataFile.LRG_REFSEQGENE, from_local)
+        )
 
     def _load_transcript_mappings_data(self, transcript_file_path: Path) -> None:
         """Load transcript mappings file to dictionaries.
@@ -99,9 +108,9 @@ class TranscriptMappings:
                         ).append(transcript)
                         self.ensembl_transcript_to_gene_symbol[transcript] = gene
                     if versioned_transcript and versioned_protein_transcript:
-                        self.ensp_to_enst[
-                            versioned_protein_transcript
-                        ] = versioned_transcript
+                        self.ensp_to_enst[versioned_protein_transcript] = (
+                            versioned_transcript
+                        )
 
     def _load_refseq_gene_symbol_data(self, lrg_refseqgene_path: Path) -> None:
         """Load data from RefSeq Gene Symbol file to dictionaries.
@@ -134,7 +143,7 @@ class TranscriptMappings:
                     if refseq_transcript and rna_transcript:
                         self.np_to_nm[refseq_transcript] = rna_transcript
 
-    def protein_transcripts(self, identifier: str) -> List[str]:
+    def protein_transcripts(self, identifier: str) -> list[str]:
         """Return a list of protein transcripts for a gene symbol.
 
         >>> from cool_seq_tool.sources import TranscriptMappings
@@ -154,7 +163,7 @@ class TranscriptMappings:
         protein_transcripts += self.refseq_protein_for_gene_symbol.get(identifier, "")
         return list(set(protein_transcripts))
 
-    def coding_dna_transcripts(self, identifier: str) -> List[str]:
+    def coding_dna_transcripts(self, identifier: str) -> list[str]:
         """Return transcripts from a coding dna refseq for a gene symbol.
 
         :param identifier: Gene identifier to find transcripts for
@@ -172,7 +181,7 @@ class TranscriptMappings:
         )
         return list(set(genomic_transcripts))
 
-    def get_gene_symbol_from_ensembl_protein(self, q: str) -> Optional[str]:
+    def get_gene_symbol_from_ensembl_protein(self, q: str) -> str | None:
         """Return the gene symbol for a Ensembl Protein.
 
         :param q: ensembl protein accession
@@ -184,7 +193,7 @@ class TranscriptMappings:
             gene_symbol = self.ensembl_protein_to_gene_symbol.get(q)
         return gene_symbol
 
-    def get_gene_symbol_from_refeq_protein(self, q: str) -> Optional[str]:
+    def get_gene_symbol_from_refeq_protein(self, q: str) -> str | None:
         """Return the gene symbol for a Refseq Protein.
 
         :param q: RefSeq protein accession
@@ -192,7 +201,7 @@ class TranscriptMappings:
         """
         return self.refseq_protein_to_gene_symbol.get(q)
 
-    def get_gene_symbol_from_refseq_rna(self, q: str) -> Optional[str]:
+    def get_gene_symbol_from_refseq_rna(self, q: str) -> str | None:
         """Return gene symbol for a Refseq RNA Transcript.
 
         :param q: RefSeq RNA transcript accession
@@ -204,7 +213,7 @@ class TranscriptMappings:
             gene_symbol = self.refseq_rna_to_gene_symbol.get(q)
         return gene_symbol
 
-    def get_gene_symbol_from_ensembl_transcript(self, q: str) -> Optional[str]:
+    def get_gene_symbol_from_ensembl_transcript(self, q: str) -> str | None:
         """Return gene symbol for an Ensembl Transcript.
 
         :param q: Ensembl transcript accession