RF2 to OWL transformation

RF2 to OWL transformation

Description

Years ago, Kent Spackman wrote a Perl script to convert RF2 into OWL. That script was not complete, nor meant to be. This project aims to improve the script and produce a guide on how a user could make the conversion (perhaps using a Perl script, perhaps something else like Python or Java). Deliverables are a well functioning script, a guidance document on how to use it and do the conversion, and a webpage that explains what it is about.

Objectives

SNOMED CT is converted to OWL for use in research projects. Some projects involve analysis of SNOMED-coding biomedical data. IHTSDO has  reputational stake in taking steps to ensure that clinical data are SNOMED-coded consistently across user groups. One step IHTSDO can take is to provide guidance to stakeholders on the technologies with which SNOMED interoperates. This proposal would result in such guidance on the conversion of SNOMED relationships to an OWL format. Once converted, the OWL representation would be used with existing ontology software tools.

The project will have a well documented audit trail that can be presented to the Management Board.

Scope

The project will focus on an RF2 snapshot - a current picture. Historic relationships, inactive concepts, metadata, etc. are out of scope. The project will focus only on stated relationships rather than inferences. 

The current script is focused on allowing classification using standard tools. Probably that should not change. The scope of RF2 is wider and that's not necessarily a problem. Only low-hanging fruit, e.g. more extensive, standardised annotations, should be allowed. Inactive content is/should not be used for classification and is thus out of scope (and is also, debatably, outside the scope of OWL).

Work Updates

UPDATE 8 Feb 2016 (provided by RTU)

RTU had a chat with Suzy today about use cases for the RF2-OWL script. She had mostly no response to her request for use cases, but these comments/issues emerged: 

  1. Some limitations had been noted regarding Protege's ability to classify more than one ontology (module?) at a time. 

  2. The latest version of the PERL script contains references to SNOMED versions prior to the release in which the script is included. The references seems to be in comments and OUT statements. This might not adversely affect the operation of the script, but it confuses novices.  

Regarding (1) RTU would like to get advice from the MAG as to whether there are limitations, and if so, what they are, and what we should do about them.

Regarding (2) we should make the comments SNOMED-version-independent, and use variables for the SNOMED version in the OUT statements so that they could refer to any release version, and not be restricted to one version only. 

 

UPDATE 9 Feb 2016 (provided by YGA)

YGA advised that the latest version of the PERL script does in fact determine the release version from the input files.  See "Relevant Documents" section below.

UPDATE 4 April 2016 (provided by HSolbrig)

Spackman Perl Transform - End User Guide

UPDATE 9 August 2016 (provided by HSolbrig)

Representation of SNOMED in OWL.v0.9

Conference Call 20 Sept 2016

Use Cases

To be provided by NLM?  See update 8 February 2016 above.

 

Requirements

  • Ultimately the script should not change from one release to another, and as a release-independent resource, it could be made permanently available rather than being distributed with each release.

Status

In Progress

Team

  • @Former user (Deleted)

  • @Peter Williams

  • @Yongsheng Gao

  • @Former user (Deleted)

  • @Former user (Deleted) 

  • @Guillermo Reynoso

  • @Daniel Karlsson (Unlicensed)

Child Pages 

Relevant Documents

  File Modified

File tls2_StatedRelationshipsToOwlKRSS_INT_general.pl

2016-Feb-09 by Peter Williams

PNG File Screen Shot 2016-02-09 at 14.23.20.png

2016-Feb-09 by Peter Williams

PDF File OWLWorkPlan.pdf Proposed approach to formalizing the "Spackman Perl representation"

2016-Apr-18 by Former user

Copyright © 2025, SNOMED International