Spackman Perl Transform - End User Guide

 This document describes the behavior of the current "Spackman Perl Script", the defacto standard for the representation of SNOMED CT in OWL. It then makes several recommendations about possible issues and how the transformation process might be improved.

Contents

Introduction

The file analyzed in this document is tls2_StatedRelationshipsToOwlKRSS_INT_20160131.pl, as distributed in the January 2016 SNOMED CT International release. The Spackman Perl Script is capable of generating OWL in either XML or Functional syntax as well as KRSS. This document focuses specifically on the OWL and, when examples are needed, uses the OWL Turtle syntax in place of the default choices above. This choice was made because of readability and greater community familiarity with Turtle vs. OWLF.

Distinguished Concept Identifiers

The transform calls out a number of special concept identifiers that are used in the transformation itself. The comment associated with the identifier list states:

# CAUTION: The values for these parameters depend on the particular release of SNOMED CT.
# Do not assume they remain stable across different releases.
# **************************************************************
# These values are valid for: 20150131, release format 2 (RF2).
# **************************************************************

The following table describes the distinguished the SNOMED CT concept codes that are used in the actual transformation process:

Namespaces used in the transformation output

The Transformation Header

The namespaces and ontology identification section is hard-coded. In particular, the module id, version URI and release date are static entries and have to be edited every release.

Issue: The elements in bold above are hard-coded. Changing the moduleId variable does not change any of the URI's above.

Issue: The output paths for OWL / OWLF / KRSS are separate. Any changes in output have to be realized the same way 3 times in a row. The OWL and OWLF output have diverged -- one gets different results for each path.

Object properties

Every active descendant of 410662002 | Concept model attribute | with the exception of 116680003|is a| is defined as an owl:ObjectProperty using the following template:

Example:

Issue: The preferred term is always "term.en-us.preferred" no matter which language is selected in the header.

Issue: The language code on the actual text is always "en", no matter what language and dialect is included in the input.

Issue: The OWL XML syntax only prints the FSN and preferred name. The OWL functional syntax includes synonyms.

Issue: The script assumes (correctly at the moment) that properties: (a) have a single parent and (b) are primitive.

Issue: The FSN is assumed to be English. This script takes the stance that FSN's are not language specific (or are English specific), which is not necessarily true.

SNOMED CT Concepts and Descriptions

Each active SNOMED CT Concept that is defined in the core module id is defined as an owl:Class using the following template:

Zero or more synonyms are emitted and (curiously) the code is designed to deal with concepts that have no FSN's.

Issue: neither Preferred nor Synonym is explicitly defined as an annotation property in the OWL XML syntax. This appears to work, but is a bit sloppy.

SNOMED CT Definitions

If the SNOMED CT is associated with an active entry in the RFW definitions file, the following is emitted:

Issue: The script assumes that there is exactly one, English language, definition. There is no connection to the Language Refset, and "en" is hard coded.

SNOMED CT Relationships

An active RF2 relationship entry as used in the RF2 stated relationship snapshot file has the following components:

The script assumes that every sourceId fits one of three situations:

1. There are no active relationship entries for sourceId -- the entry is treated as a root node and the definition is considered to be complete.
   
   
   

2. There is exactly one active relationship entry for sourceId -- the typeId for the entry is assumed to be 116680003|is a| (unchecked!) and the following OWL is emitted:
   rdfs:subClassOf sctid:{destinationId} .

3. There are two or more active relationship entries for sourceId. The processing rules for this situation are defined in the next section.

Processing Rules for 2 or more relationship entries

If there are two or more active relationship entries for a given subject SCTID, the definition status for concept determines whether the definition will be in the form:

if the definition status is 900000000000073002|Defined| and:

otherwise.

The intersection is filled out as follows:

1. For every relationship entry with typeId = 116680003|is a| , output the URI of the corresponding destinationId.

2. For every unique non-zero relationshipGroup entry, emit:
   [ a owl:Restriction
owl:onProperty sctid:609096000 ;
owl:someValuesFrom [ a owl:Class ;
owl:intersectionOf ( ...) ] ]
   Followed by:
   [ a owl:Restriction ;
owl:onProperty sctid:{typeId};
owl:someValuesFrom sctid:{destinationId} ]
   for every row in the group. Note: owl:intersectionOf is not emitted if there is only one row in a given group.
   
   
   

3. For each zero relationshipGroup entry with a typeId other than 116680003|is a|:
   if typeId is one of the "never grouped" types emit:
   [ a owl:Restriction ;
owl:onProperty sctid:{typeId};
owl:someValuesFrom sctid:{destinationId} ]
   otherwise:
   [ a owl:Restriction ;
owl:onProperty sctid:609096000 ;
owl:someValuesFrom [ a owl:Restriction ;
owl:onProperty sctid:{typeId};
owl:someValuesFrom sctid:{destinationId} ]]

Examples:

A Root concept

sctid:138875005 a owl:Class ;

rdfs:label "SNOMED CT Concept (SNOMED RT+CTV3)"@en ;

sctf:Description.term.en-us.preferred "SNOMED CT Concept"@en ;

sctf:Description.term.en-us.synonym "SNOMED CT has been created by combining SNOMED RT and a computer-based nomenclature and classification known as Read Codes Version 3, which was created on behalf of the U.K. Department of Health."@en,

"SNOMED Clinical Terms version: 20160131 [R] (January 2016 Release)"@en .

A concept with a single parent

sctid:14120002 a owl:Class ;

rdfs:label "Rodenticide (substance)"@en ;

sctf:Description.term.en-us.preferred "Rodenticide"@en ;

rdfs:subClassOf sctid:59545008 .

A primitive concept with two parents

sctid:108003 a owl:Class ;

rdfs:label "Entire condylar emissary vein (body structure)"@en ;

sctf:Description.term.en-us.preferred "Entire condylar emissary vein"@en ;

sctf:Description.term.en-us.synonym "Condylar emissary vein"@en ;

rdfs:subClassOf [ a owl:Class ;

owl:intersectionOf ( sctid:59191008 sctid:154631009 ) ] .

A fully defined concept with one parent and one non-zero role group

sctid:74400008 a owl:Class ;

rdfs:label "Appendicitis (disorder)"@en ;

sctf:Description.term.en-us.preferred "Appendicitis"@en ;

owl:equivalentClass [ a owl:Class ;

owl:intersectionOf ( sctid:18526009 [ a owl:Restriction ;

owl:onProperty sctid:609096000 ;

owl:someValuesFrom [ a owl:Class ;

owl:intersectionOf ( [ a owl:Restriction ;

owl:onProperty sctid:116676008 ;

owl:someValuesFrom sctid:23583003 ] [ a owl:Restriction ;

owl:onProperty sctid:363698007 ;

owl:someValuesFrom sctid:66754008 ] ) ] ] ) ] .

A concept with both zero and non-zero role groups

Relationship file entries:

sourceId destinationId relationshipGroup typeId

425630003 110979008 0 116680003

425630003 111189002 0 116680003

425630003 105590001 0 246075003

425630003 400195000 0 42752001

425630003 424124008 0 263502005

425630003 472963003 0 370135005

425630003 39937001 1 363698007

425630003 4532008 1 116676008

Compositional grammar definition:

425630003| Acute irritant contact dermatitis | =

111189002| Acute contact dermatitis| +

110979008| Irritant contact dermatitis|:

42752001| Due to |=400195000| Contact hypersensitivity reaction|,

246075003| Causative agent|=105590001| Substance|,

263502005| Clinical course|=424124008| Sudden onset AND/OR short duration|,

370135005| Pathological process|=472963003| Hypersensitivity process|,

{ 363698007| Finding site |=39937001| Skin structure|,

116676008| Associated morphology |=4532008| Acute inflammation | }

OWL:

scid:425630003 a owl:Class ;

rdfs:label "Acute irritant contact dermatitis (disorder)"@en ;

sctf:Description.term.en-us.preferred "Acute irritant contact dermatitis"@en ;

owl:equivalentClass [ a owl:Class ;

owl:intersectionOf ( sctid:110979008 sctid:111189002 [ a owl:Restriction ;

owl:onProperty sctid:609096000 ;

owl:someValuesFrom [ a owl:Restriction ;

owl:onProperty sctid:246075003 ;

owl:someValuesFrom sctid:105590001 ] ] [ a owl:Restriction ;

owl:onProperty sctid:609096000 ;

owl:someValuesFrom [ a owl:Restriction ;

owl:onProperty sctid:42752001 ;

owl:onProperty sctid:400195000 ] ] [ a owl:Restriction ;

owl:onProperty sctid:609096000 ;

owl:someValuesFrom [ a owl:Class ;

owl:intersectionOf ( [ a owl:Restriction ;

owl:onProperty sctid:363698007;

owl:someValuesFrom sctid:39937001> ] [ a owl:Restriction ;

owl:onProperty sctid:116676008 ;

owl:someValuesFrom sctid:4532008 ] ) ] ] ) ] .

A concept with mixed relationshipGroup zeros

sourceId destinationId relationshipGroup typeId

10243007 385101003 0 411116001 never grouped

10243007 387253001 0 127489000 never grouped

10243007 387350000 0 127489000 never grouped

10243007 420081005 0 116680003 One parent

sctid:10243007 a owl:Class ;

rdfs:label "Benzoic and salicylic acid ointment (product)"@en ;

sctf:Description.term.en-us.preferred "Benzoic and salicylic acid ointment"@en ;

sctf:Description.term.en-us.synonym "Whitfield's ointment"@en ;

rdfs:subClassOf [ a owl:Class ;

owl:intersectionOf ( sctid:420081005

[ a owl:Restriction ;

owl:onProperty sctid:411116001 ;

owl:someValuesFrom sctid:385101003 ]

[ a owl:Restriction ;

owl:onProperty sctid:127489000 ;

owl:someValuesFrom sctid:387253001 ]

[ a owl:Restriction ;

owl:onProperty sctid:127489000 ;

owl:someValuesFrom sctid:387350000 ] ) ] .

Notes and Observations:

1. There are a number of elements that are hard-coded into the script, including dates, the ontology identifier itself, etc. This makes the script less than useful for anything but the SNOMED CT Core module.

2. Knowledge about "never grouped" concepts are in the script but not in the RF2 release. This tacit knowledge should be made explicit in the RF2 release so that this sort of editing is not required and the script can be applied to other modules/extensions.

3. Knowledge about "right identity" (property chain) concepts are in the script but not in the RF2 release. This tacit knowledge should be made explicit in the RF2 release so that this sort of editing is not required and the script can be applied to other modules/extensions.

4. The script makes a number of assumptions about the consistency and integrity of the RF2 release. It also makes assumptions about which aspects of the RF2 release are actually in use (e.g. "All", one active FSN per concept, at most one definition per concept which is assumed to be in english, etc.)

5. The script is not designed to address modules other than the root module. The root module is hard-coded and the module dependency file is ignored, meaning that no import statements are included in the output.

6. The script is currently us-english centric and requires modification to emit other dialects of languages. It is also not equipped to emit all language variants.

7. The script includes unnecessary owl:intersectionOf assertions on rdfs:subClassOf entries. This makes the output less readable and non-standard.

8. The script should use skos:definition, skos:prefLabel and skos:altLabel instead of defining special properties in the sctp and sctf namespace.

9. The script needs to address some of the currently unused features in the RF2, including the "qualifying" and "additional" relationship types as well as the "All" modifier. If these features are not going to be used, they should be removed so that extension authors will not be tempted.

10. The script needs to address the elements in the "Unapproved attribute" branch.