diff --git a/acmc/main.py b/acmc/main.py
index 29f5aea00f8c1c5384ed48970cce9204cdda45b1..02ad902c9a1de0dcaa50797b710b3b7828109bb5 100644
--- a/acmc/main.py
+++ b/acmc/main.py
@@ -58,7 +58,13 @@ def _phen_validate(args: argparse.Namespace):
def _phen_map(args: argparse.Namespace):
"""Handle the `phen map` command."""
- phen.map(args.phen_dir, args.target_coding, args.not_translate, args.no_metadata)
+ phen.map(
+ args.phen_dir,
+ args.target_coding,
+ args.not_translate,
+ args.no_metadata,
+ args.do_reverse_translate,
+ )
def _phen_export(args: argparse.Namespace):
@@ -235,6 +241,12 @@ def main():
default=False,
help="(Optional) Prevent copying of metadata columns to output.",
)
+ phen_map_parser.add_argument(
+ "--do-reverse-translate",
+ action="store_true",
+ default=False,
+ help="(Optional) Enable reversing one directional mappings. WARNING goes against NHS TRUD guidelines.",
+ )
phen_map_parser.set_defaults(func=_phen_map)
# phen export
diff --git a/acmc/phen.py b/acmc/phen.py
index 01f417faa03504b523423bad0982faa14a264973..64f6b4983db7a362a9e0f225686877fb4f9b41a8 100644
--- a/acmc/phen.py
+++ b/acmc/phen.py
@@ -666,6 +666,7 @@ def translate_codes(
target_code_type: str,
concept_name: str,
not_translate: bool,
+ do_reverse_translate: bool,
) -> pd.DataFrame:
"""Translates each source code type the source coding list into a target type and returns all conversions as a concept set"""
@@ -695,23 +696,15 @@ def translate_codes(
filename = f"{source_code_type}_to_{target_code_type}.parquet"
map_path = trud.PROCESSED_PATH / filename
+ filename_reversed = f"{target_code_type}_to_{source_code_type}.parquet"
+ map_path_reversed = trud.PROCESSED_PATH / filename_reversed
+
# do the mapping if it exists
if map_path.exists():
- # get mapping
- df_map = pd.read_parquet(map_path)
-
- # do mapping
- translated_df = pd.merge(
- source_df[source_code_type], df_map, how="left"
- )
-
- # normalise the output
- translated_df.columns = pd.Index(["SOURCE_CONCEPT", "CONCEPT"])
- translated_df["SOURCE_CONCEPT_TYPE"] = source_code_type
-
- # add to list of codes
- codes = pd.concat([codes, translated_df])
-
+ codes = _translate_codes(map_path, source_df, source_code_type, codes)
+ # otherwise do reverse mapping if enabled and it exists
+ elif do_reverse_translate and map_path_reversed.exists():
+ codes = _translate_codes(map_path_reversed, source_df, source_code_type, codes)
else:
_logger.warning(
f"No mapping from {source_code_type} to {target_code_type}, file {str(map_path.resolve())} does not exist"
@@ -728,6 +721,23 @@ def translate_codes(
return codes
+def _translate_codes(map_path, source_df, source_code_type, codes) -> pd.DataFrame:
+ # get mapping
+ df_map = pd.read_parquet(map_path)
+
+ # do mapping
+ translated_df = pd.merge(source_df[source_code_type], df_map, how="left")
+
+ # normalise the output
+ translated_df.columns = pd.Index(["SOURCE_CONCEPT", "CONCEPT"])
+ translated_df["SOURCE_CONCEPT_TYPE"] = source_code_type
+
+ # add to list of codes
+ codes = pd.concat([codes, translated_df])
+
+ return codes
+
+
def _write_code_errors(code_errors: list, code_errors_path: Path):
err_df = pd.DataFrame(
[
@@ -786,7 +796,13 @@ def write_vocab_version(phen_path: Path):
)
-def map(phen_dir: str, target_code_type: str, not_translate: bool, no_metadata: bool):
+def map(
+ phen_dir: str,
+ target_code_type: str,
+ not_translate: bool,
+ no_metadata: bool,
+ do_reverse_translate: bool,
+):
_logger.info(f"Processing phenotype: {phen_dir}")
# Validate configuration
@@ -811,11 +827,23 @@ def map(phen_dir: str, target_code_type: str, not_translate: bool, no_metadata:
if target_code_type is not None:
_map_target_code_type(
- phen_path, phenotype, target_code_type, not_translate, no_metadata
+ phen_path,
+ phenotype,
+ target_code_type,
+ not_translate,
+ no_metadata,
+ do_reverse_translate,
)
else:
for t in phenotype["map"]:
- _map_target_code_type(phen_path, phenotype, t, not_translate, no_metadata)
+ _map_target_code_type(
+ phen_path,
+ phenotype,
+ t,
+ not_translate,
+ no_metadata,
+ do_reverse_translate,
+ )
_logger.info(f"Phenotype processed successfully")
@@ -826,6 +854,7 @@ def _map_target_code_type(
target_code_type: str,
not_translate: bool,
no_metadata: bool,
+ do_reverse_translate: bool,
):
_logger.debug(f"Target coding format: {target_code_type}")
concepts_path = phen_path / CONCEPTS_DIR
@@ -894,6 +923,7 @@ def _map_target_code_type(
target_code_type=target_code_type,
concept_name=concept_set_name,
not_translate=not_translate,
+ do_reverse_translate=do_reverse_translate,
)
trans_out = add_metadata(
codes=trans_out,
@@ -908,6 +938,7 @@ def _map_target_code_type(
target_code_type=target_code_type,
concept_name=concept_set_name,
not_translate=not_translate,
+ do_reverse_translate=do_reverse_translate,
)
trans_out = add_metadata(
codes=trans_out,
diff --git a/docs/cli.md b/docs/cli.md
index 396f5291183990337ed7d5ce98192673da93ca4c..6001ec7dc2a805c170f070c951c9583bf4cb4184 100644
--- a/docs/cli.md
+++ b/docs/cli.md
@@ -112,7 +112,8 @@ The `phen` command is used phenotype-related operations.
- `-t`, `--target-coding`: (Optional) Specify the target coding (e.g., `read2`, `read3`, `icd10`, `snomed`, `opcs4`).
- `-d`, `--phen-dir`: (Optional) Local phenotype workspace directory (default is ./workspace/phen).
- `--not-translate`: (Optional) Prevent any phenotype translation using NHS TRUD vocabularies. Therefore only concepts in already in the traget coding will be mapped.
- - `--no-metadata`: (Optional) Prevent copying of metadata columns to output.
+ - `--no-metadata`: (Optional) Prevent copying of metadata columns to output.
+ - `--do-reverse-translate`: (Optional) Enable reversing one directional mappings. **WARNING** goes against NHS TRUD guidelines.
- **Publish Phenotype Configuration**