New Feature: SNOMED::ICD10CM Mapping Support

- Added feature to allow for conversion of these premade mappings provided by SNOMED into SSSOM format.

General updates
- cli.py: Reorganized SSSOM_READ_FORMATS: Top half are plain data formats, and bottom half are special-case formats. Both halves of the list are alphabetically sorted.

Temp updates
- Changed some relative imports to absolute imports, in order to speed up development and make debugging easier. It is possible that this could be a good permanent change too, though.

Author: joeflack4

sssom/parsers.py

@@ -5,6 +5,7 @@
 import re
 import typing
 from collections import Counter
+from dateutil import parser as date_parser
 from pathlib import Path
 from typing import Any, Callable, Dict, List, Optional, Set, TextIO, Tuple, Union, cast
 from urllib.request import urlopen
@@ -24,7 +25,7 @@
     add_built_in_prefixes_to_prefix_map,
     get_default_metadata,
 )
-from .sssom_datamodel import Mapping, MappingSet
+from .sssom_datamodel import Mapping, MappingSet, MatchTypeEnum
 from .sssom_document import MappingSetDocument
 from .typehints import Metadata, MetadataType, PrefixMap
 from .util import (
@@ -140,6 +141,24 @@ def read_obographs_json(
     )
 
 
+def read_snomed_icd10cm_map_tsv(
+    file_path: str,
+    prefix_map: Dict[str, str] = None,
+    meta: Dict[str, str] = None,
+) -> MappingSetDataFrame:
+    """Parse special SNOMED ICD10CM mapping file and translates it into a MappingSetDataFrame.
+
+    :param file_path: The path to the obographs file
+    :param prefix_map: an optional prefix map
+    :param meta: an optional dictionary of metadata elements
+    :return: A SSSOM MappingSetDataFrame
+    """
+    raise_for_bad_path(file_path)
+    df = read_pandas(file_path)
+    df2 = from_snomed_icd10cm_map_tsv(df, prefix_map=prefix_map, meta=meta)
+    return df2
+
+
 def _get_prefix_map_and_metadata(
     prefix_map: Optional[PrefixMap] = None, meta: Optional[MetadataType] = None
 ) -> Metadata:
@@ -499,6 +518,144 @@ def from_obographs(
     return to_mapping_set_dataframe(mdoc)
 
 
+def from_snomed_icd10cm_map_tsv(
+    df: pd.DataFrame,
+    prefix_map: Optional[PrefixMap] = None,
+    meta: Optional[MetadataType] = None,
+) -> MappingSetDataFrame:
+    """Convert a snomed_icd10cm_map dataframe to a MappingSetDataFrame.
+
+    :param df: A mappings dataframe
+    :param prefix_map: A prefix map
+    :param meta: A metadata dictionary
+    :return: MappingSetDataFrame
+
+    # Field descriptions
+    # - Taken from: doc_Icd10cmMapReleaseNotes_Current-en-US_US1000124_20210901.pdf
+    FIELD,DATA_TYPE,PURPOSE,Joe's comments
+    - id,UUID,A 128 bit unsigned integer, uniquely identifying the map record,
+    - effectiveTime,Time,Specifies the inclusive date at which this change becomes effective.,
+    - active,Boolean,Specifies whether the member’s state was active (=1) or inactive (=0) from the nominal release date
+     specified by the effectiveTime field.,
+    - moduleId,SctId,Identifies the member version’s module. Set to a child of 900000000000443000|Module| within the
+    metadata hierarchy.,The only value in the entire set is '5991000124107', which has label 'SNOMED CT to ICD-10-CM
+    rule-based mapping module' (
+    https://www.findacode.com/snomed/5991000124107--snomed-ct-to-icd-10-cm-rule-based-mapping-module.html).
+    - refSetId,SctId,Set to one of the children of the |Complex map type| concept in the metadata hierarchy.,The only
+    value in the entire set is '5991000124107', which has label 'ICD-10-CM complex map reference set' (
+    https://www.findacode.com/snomed/6011000124106--icd-10-cm-complex-map-reference-set.html).
+    - referencedComponentId,SctId,The SNOMED CT source concept ID that is the subject of the map record.,
+    - mapGroup,Integer,An integer identifying a grouping of complex map records which will designate one map target at
+    the time of map rule evaluation. Source concepts that require two map targets for classification will have two sets
+    of map groups.,
+    - mapPriority,Integer,Within a map group, the mapPriority specifies the order in which complex map records should be
+    evaluated to determine the correct map target.,
+    - mapRule,String,A machine-readable rule, (evaluating to either ‘true’ or ‘false’ at run-time) that indicates
+    whether this map record should be selected within its map group.,
+    - mapAdvice,String,Human-readable advice that may be employed by the software vendor to give an end-user advice on
+    selection of the appropriate target code. This includes a) a summary statement of the map rule logic, b) a statement
+    of any limitations of the map record and c) additional classification guidance for the coding professional.,
+    - mapTarget,String,The target ICD-10 classification code of the map record.,
+    - correlationId,SctId,A child of |Map correlation value| in the metadata hierarchy, identifying the correlation
+    between the SNOMED CT concept and the target code.,
+    - mapCategoryId,SctId,Identifies the SNOMED CT concept in the metadata hierarchy which is the MapCategory for the
+    associated map record. This is a subtype of 447634004 |ICD-10 Map Category value|.,
+    """
+    # https://www.findacode.com/snomed/447561005--snomed-ct-source-code-to-target-map-correlation-not-specified.html
+    match_type_snomed_unspecified_id = 447561005
+    prefix_map = _ensure_prefix_map(prefix_map)
+    ms = _init_mapping_set(meta)
+
+    mlist: List[Mapping] = []
+    for _, row in df.iterrows():
+        mdict = {
+            'subject_id': f'SNOMED:{row["referencedComponentId"]}',
+            'subject_label': row['referencedComponentName'],
+
+            # 'predicate_id': 'skos:exactMatch',
+            # - mapCategoryId: can use for mapping predicate? Or is correlationId more suitable?
+            #   or is there a SKOS predicate I can map to in case where predicate is unknown? I think most of these
+            #   mappings are attempts at exact matches, but I can't be sure (at least not without using these fields
+            #   to determine: mapGroup, mapPriority, mapRule, mapAdvice).
+            # mapCategoryId,mapCategoryName: Only these in set: 447637006 "MAP SOURCE CONCEPT IS PROPERLY CLASSIFIED",
+            #   447638001 "MAP SOURCE CONCEPT CANNOT BE CLASSIFIED WITH AVAILABLE DATA",
+            #   447639009 "MAP OF SOURCE CONCEPT IS CONTEXT DEPENDENT"
+            # 'predicate_modifier': '???',
+            #   Description: Modifier for negating the prediate. See https://github.com/mapping-commons/sssom/issues/40
+            #   Range: PredicateModifierEnum: (joe: only lists 'Not' as an option)
+            #   Example: Not Negates the predicate, see documentation of predicate_modifier_enum
+            # - predicate_id <- mapAdvice?
+            # - predicate_modifier <- mapAdvice?
+            #   mapAdvice: Pipe-delimited qualifiers. Ex:
+            #   "ALWAYS Q71.30 | CONSIDER LATERALITY SPECIFICATION"
+            #   "IF LISSENCEPHALY TYPE 3 FAMILIAL FETAL AKINESIA SEQUENCE SYNDROME CHOOSE Q04.3 | MAP OF SOURCE CONCEPT
+            #   IS CONTEXT DEPENDENT"
+            #   "MAP SOURCE CONCEPT CANNOT BE CLASSIFIED WITH AVAILABLE DATA"
+            'predicate_id': f'SNOMED:{row["mapCategoryId"]}',
+            'predicate_label': row['mapCategoryName'],
+
+            'object_id': f'ICD10CM:{row["mapTarget"]}',
+            'object_label': row['mapTargetName'],
+
+            # match_type <- mapRule?
+            #   ex: TRUE: when "ALWAYS <code>" is in pipe-delimited list in mapAdvice, this always shows TRUE. Does this
+            #       mean I could use skos:exactMatch in these cases?
+            # match_type <- correlationId?: This may look redundant, but I want to be explicit. In officially downloaded
+            #   SNOMED mappings, all of them had correlationId of 447561005, which also happens to be 'unspecified'.
+            #   If correlationId is indeed more appropriate for predicate_id, then I don't think there is a representative
+            #   field for 'match_type'.
+            'match_type': MatchTypeEnum('Unspecified') if row['correlationId'] == match_type_snomed_unspecified_id \
+                else  MatchTypeEnum('Unspecified'),
+
+            'mapping_date': date_parser.parse(str(row['effectiveTime'])).date(),
+            'other': '|'.join([f'{k}={str(row[k])}' for k in [
+                'id',
+                'active',
+                'moduleId',
+                'refsetId',
+                'mapGroup',
+                'mapPriority',
+                'mapRule',
+                'mapAdvice',
+            ]]),
+
+            # More fields (https://mapping-commons.github.io/sssom/Mapping/):
+            # - subject_category: absent
+            # - author_id: can this be "SNOMED"?
+            # - author_label: can this be "SNOMED"?
+            # - reviewer_id: can this be "SNOMED"?
+            # - reviewer_label: can this be "SNOMED"?
+            # - creator_id: can this be "SNOMED"?
+            # - creator_label: can this be "SNOMED"?
+            # - license: Is this something that can be determined?
+            # - subject_source: URL of some official page for SNOMED version used?
+            # - subject_source_version: Is this knowable?
+            # - objectCategory <= mapRule?
+            #   mapRule: ex: TRUE: when "ALWAYS <code>" is in pipe-delimited list in mapAdvice, this always shows TRUE.
+            #     Does this mean I could use skos:exactMatch in these cases?
+            #     object_category:
+            #   objectCategory:
+            #     Description: The conceptual category to which the subject belongs to. This can be a string denoting
+            #     the category or a term from a controlled vocabulary.
+            #     Example: UBERON:0001062 (The CURIE of the Uberon term for "anatomical entity".)
+            # - object_source: URL of some official page for ICD10CM version used?
+            # - object_source_version: would this be "10CM" as in "ICD10CM"? Or something else? Or nothing?
+            # - mapping_provider: can this be "SNOMED"?
+            # - mapping_cardinality: Could I determine 1:1 or 1:many or many:1 based on:
+            #   mapGroup, mapPriority, mapRule, mapAdvice?
+            # - match_term_type: What is this?
+            # - see_also: Should this be a URL to the SNOMED term?
+            # - comment: Description: Free text field containing either curator notes or text generated by tool providing
+            #   additional informative information.
+        }
+        mlist.append(_prepare_mapping(Mapping(**mdict)))
+
+    ms.mappings = mlist
+    _set_metadata_in_mapping_set(mapping_set=ms, metadata=meta)
+    doc = MappingSetDocument(mapping_set=ms, prefix_map=prefix_map)
+    return to_mapping_set_dataframe(doc)
+
+
 # All from_* take as an input a python object (data frame, json, etc) and return a MappingSetDataFrame
 # All read_* take as an input a a file handle and return a MappingSetDataFrame (usually wrapping a from_* method)
 
@@ -523,6 +680,9 @@ def get_parsing_function(input_format: Optional[str], filename: str) -> Callable
         return read_alignment_xml
     elif input_format == "obographs-json":
         return read_obographs_json
+    elif input_format == "snomed-icd10cm-map-tsv":
+        return read_snomed_icd10cm_map_tsv
+
     else:
         raise Exception(f"Unknown input format: {input_format}")
 

sssom/util.py

@@ -43,12 +43,13 @@
 PREFIX_MAP_KEY = "curie_map"
 
 SSSOM_READ_FORMATS = [
-    "tsv",
-    "rdf",
+    "json",
     "owl",
+    "rdf",
+    "tsv",
     "alignment-api-xml",
     "obographs-json",
-    "json",
+    "snomed-icd10cm-map-tsv"
 ]
 SSSOM_EXPORT_FORMATS = ["tsv", "rdf", "owl", "json"]