PaperlayerView exact revision on arXiv

Source-linked AI summary

A Simple Standard for Sharing Ontological Mappings (SSSOM)

Nicolas Matentzoglu, James P. Balhoff, Susan M. Bello, Chris Bizon, Matthew Brush, Tiffany J. Callahan, Christopher G Chute, William D. Duncan, Chris T. Evelo, Davera Gabriel, John Graybeal, Alasdair Gray, Benjamin M. Gyori, Melissa Haendel, Henriette Harmse, Nomi L. Harris, Ian Harrow, Harshad Hegde, Amelia L. Hoyt, Charles T. Hoyt, Dazhi Jiao, Ernesto Jiménez-Ruiz, Simon Jupp, Hyeongsik Kim, Sebastian Koehler, Thomas Liener, Qinqin Long, James Malone, James A. McLaughlin, Julie A. McMurry, Sierra Moxon, Monica C. Munoz-Torres, David Osumi-Sutherland, James A. Overton, Bjoern Peters, Tim Putman, Núria Queralt-Rosinach, Kent Shefchek, Harold Solbrig, Anne Thessen, Tania Tudorache, Nicole Vasilevsky, Alex H. Wagner, Christopher J. Mungall

arXiv:2112.07051v1cs.DB

TL;DR

Mapping sets often lack transparent precision, accuracy, completeness, and reuse metadata, with no widely agreed exchange standard. SSSOM combines a flat metadata model with a simple tabular format and community-driven governance; its use cases include semantic access to mappings and Mondo’s provenance-bearing exports. The standard’s scope excludes mappings that depend on global context.

Abstract

from arXiv · show

Despite progress in the development of standards for describing and exchanging scientific information, the lack of easy-to-use standards for mapping between different representations of the same or similar objects in different databases poses a major impediment to data integration and interoperability. Mappings often lack the metadata needed to be correctly interpreted and applied. For example, are two terms equivalent or merely related? Are they narrow or broad matches? Are they associated in some other way? Such relationships between the mapped terms are often not documented, leading to incorrect assumptions and making them hard to use in scenarios that require a high degree of precision (such as diagnostics or risk prediction). Also, the lack of descriptions of how mappings were done makes it hard to combine and reconcile mappings, particularly curated and automated ones. The Simple Standard for Sharing Ontological Mappings (SSSOM) addresses these problems by: 1. Introducing a machine-readable and extensible vocabulary to describe metadata that makes imprecision, inaccuracy and incompleteness in mappings explicit. 2. Defining an easy to use table-based format that can be integrated into existing data science pipelines without the need to parse or query ontologies, and that integrates seamlessly with Linked Data standards. 3. Implementing open and community-driven collaborative workflows designed to evolve the standard continuously to address changing requirements and mapping practices. 4. Providing reference tools and software libraries for working with the standard. In this paper, we present the SSSOM standard, describe several use cases, and survey some existing work on standardizing the exchange of mappings, with the goal of making mappings Findable, Accessible, Interoperable, and Reusable (FAIR). The SSSOM specification is at http://w3id.org/sssom/spec.

Introduction

SSSOM provides a standard and software ecosystem for exchanging, validating, and summarizing mappings with interpretable relationship and provenance metadata. The section also describes an EOSC-Life use case combining SSSOM with a FAIR Data Point for semantically informed conversion from SNOMED CT to Orphacodes.

Introduction

  • SSSOM aims to make mappings’ imprecision, inaccuracy, and incompleteness explicit while supporting FAIR sharing, data-science toolchains, community governance, and diverse mapping types.The standard responds to the risks of missing or inaccurate mappings and the limitations of ad-hoc, often single-use exchange formats.Source: Introduction · p. 3, Desired features of a standard for… · p. 6

Data model

  • SSSOM represents pairwise mappings with required subject, object, predicate, and match-type fields, plus optional metadata for match methods and provenance.Version 0.9 defines up to 38 metadata slots per mapping; match types include lexical, logical, human-curated, semantic-similarity, and complex.Source: Data model · p. 7, Predicates · p. 8, Provenance · p. 9
  • SSSOM’s native TSV format keeps mappings and metadata in a simple table that general data-science tools can read without specialized software.Mapping-set metadata can be embedded in a commented YAML header or supplied externally; embedded headers are promoted to reduce provenance loss when sharing files.Source: SSSOM TSV format: a simple tabular… · p. 10

Sustainability: Collaborative workflows and governance

  • SSSOM schema changes use GitHub issues and reviewed pull requests, with core-team approval required for changes that are not backwards compatible.The project uses open-source collaborative workflows, including issue templates, reviews, and continuous integration.Source: Sustainability: Collaborative… · p. 11

The SSSOM software ecosystem

  • SSSOM’s software ecosystem includes tools to convert, validate, merge, query, and generate mapping tables, alongside an early-stage vision for public mapping commons.sssom-py supports imports and exports across formats, while rdf-matcher exports mapping rules and metadata such as confidence and match fields.Source: The SSSOM software ecosystem · p. 11
  • Mondo uses probability-scored mapping candidates with precise predicates in k-BOOM to identify likely correct mappings, then distributes mappings as SSSOM tables with provenance.The tables enable downstream users to use mappings that were previously distributed mainly as ontology axioms or annotations.Source: Use Case 1: Harmonizing disease… · p. 12
  • OxO mapping walks can return obsolete or merely related terms, illustrating why mappings need explicit precision metadata and rules that users can inspect.SSSOM is intended to enable precise cross-walks and provide a shared way to represent mapping rules, which OxO could then implement.Source: Use Case 2: Browsing and cross-walking… · pp. 12–13
  • The National Microbiome Data Collaborative aligns source metadata schemas and sample vocabularies, including the GOLD environmental path vocabulary and ENVO.The alignment spans GOLD, NCBI, the NMDC schema, and the MIxS standard.Source: Use Case 3: National Microbiome Data… · p. 13
  • SSSOM mapping metadata lets NMDC summarize and validate mappings, including their predicates and whether they were expert-curated or obtained from a specific source.Most NMDC mappings are exact matches, with a small handful of related matches.Source: Use Case 3: National Microbiome Data… · p. 13
  • In the EOSC-Life use case, a FAIR Data Point exposes mappings for converting SNOMED CT disease terms to Orphacodes in the European Osteogenesis imperfecta registry.The scenario involves a data steward registering patient information when partners use different disease vocabularies.Source: Use Case 4: Finding and using mappings… · p. 13
  • SSSOM mappings carry richer metadata, such as match precision, enabling Alice to retrieve mappings by semantics and accurately convert specified subjects to mapped objects.The use case contrasts this with mapping systems that provide only simple equivalence between objects of the same subject.Source: Use Case 4: Finding and using mappings… · pp. 13–14

Related work

The related-work survey reviews mapping formats, informal efforts, and services, comparing their metadata, provenance, interoperability, and reuse support with SSSOM (p19).

Standard formats for capturing mappings

  • SSSOM adds mapping metadata and provenance to a simple exchange format; RDF Alignment and EDOAL are supported by the Alignment API, while MELT may add SSSOM support (40, 44–46).The ontology-matching community is considering SSSOM as an additional exchange format (40–46).Source: Standard formats for capturing mappings · p. 14
  • VoID linksets and BridgeDb offer dataset links and stacked mappings, but VoID cannot place multiple subjects or objects in one linkset file (47–51).The authors discuss integrating VoID linksets and BridgeDb mapping vocabulary with SSSOM mapping sets.Source: Standard formats for capturing mappings · p. 14
  • CTS2 captures many SSSOM-like metadata elements but supports complex mappings and service behavior, whereas DOL focuses on distributed semantics; CTS2 is more complex and clinically oriented (53–55).SNOMED CT mappings need not carry extensive creation provenance and are built locally; SSSOM underlies publicly available Mapping Commons projects and is not tied to one system (56).Source: Standard formats for capturing mappings · p. 15
  • UMLS maps 214 vocabularies with automated and curated methods, but its model lacks many SSSOM metadata elements and a public collaborative workflow for new metadata or formal format mappings (57–59).Some provenance can be captured with Dublin Core, OWL, PROV-O, PAV, and MQV; alignment of SSSOM properties with these vocabularies remains in progress (3, 52).Source: Standard formats for capturing mappings · pp. 15–16

Informal approaches for capturing mappings

  • OBO xrefs are semantically ambiguous across ontologies, and analysis of about a million references found their unclear semantics made reuse impractical or impossible (60, 61).The authors agree with that assessment.Source: Informal approaches for capturing… · p. 16
  • SSSOM incorporates most metadata elements from a BioHackathon survey, while SEMAF shares its Semantic Web approach but its implementation and metadata remain unsettled; OMOP2OBO’s flat-file-to-RDF/OWL conversion is being aligned with SSSOM (62–67).The BioHackathon survey did not produce a formal specification.Source: Informal approaches for capturing… · p. 16

Mapping services and tools

  • OxO is being extended to support SSSOM natively, and BioPortal intends to add SSSOM support and a SPARQL endpoint (68–70).OxO’s planned support would let it ingest any SSSOM datasource; BioPortal mappings include their timestamp, creation process, relationship, and type.Source: Mapping services and tools · p. 17
  • Biomappings distributes curated and predicted mappings in SSSOM under CC0, while Paxo and OxO alignments remain limited to CSV unless converted (71, 76–77).Biomappings uses automated integrity checks and a web curation interface; some Paxo/OxO alignments have already been converted to SSSOM (71, 74–77).Source: Mapping services and tools · p. 17

Discussion and limitations of the approach

SSSOM’s flat design trades expressivity for simplicity, leaving some context-dependent, complex-rule, modified, and complex mappings difficult or impossible to represent. The authors describe workarounds for some cases while accepting others as limitations.

Discussion and limitations of the approach

  • Context-dependent mappings are not globally scoped; relation-specific context is supported but can hide assumptions in an ontology, while complex expressions remain under discussion.The authors say many use cases can be captured without complex expressions.Source: Discussion and limitations of the… · p. 18
  • A flat model cannot readily encode multiple match types for one mapping; SSSOM’s workaround is one row per mapping rule.The rdf-matcher example would produce two rows for matches supported by synonym and primary-label lexical matches.Source: Discussion and limitations of the… · p. 18
  • Adding metadata can change a mapping table’s semantics, and consumers may overlook predicate modifiers such as negation.The authors chose a predicate_modifier field to preserve simplicity, accepting the risk that consumers may not notice modified mappings.Source: Discussion and limitations of the… · pp. 18–19
  • Ignoring preprocessing metadata can mislead interpretation: stripping non-alphabetic characters could make the distinct concepts “Alzheimer 2” and “Alzheimer 3” appear matched.The authors give this as another violation of idempotency.Source: Discussion and limitations of the… · p. 19
  • SSSOM does not currently support complex mappings, where a subject or object is an expression involving more than one term.The authors postponed support to protect the simple metadata model and are considering an extension.Source: Discussion and limitations of the… · p. 19

Conclusions and Future Work

Mapping standards often lack metadata and shared workflows for transparent, reusable mappings, as well as a simple standardized tabular representation. SSSOM addresses these gaps with an extensible metadata vocabulary, community-driven governance, and a table format, while the authors also emphasize shared practices and future training and tool development.

Conclusions and Future Work

  • Existing mapping standards often lack vocabulary to make imprecision, inaccuracy, and incompleteness explicit, open community workflows, and standardized tabular representation.The authors identify these as shortcomings of prior attempts to standardize mappings.Source: Conclusions and Future Work · p. 19
  • SSSOM combines rich mapping metadata, community-driven governance, and a simple tabular format for disseminating mappings in data-science workflows.The standard is presented as addressing the identified shortcomings.Source: Conclusions and Future Work · pp. 19–20
  • A simple standardized format is a prerequisite for high-quality mappings and sharing, but improving mapping quality also requires shared best practices.The authors describe collaboration with OpenTargets, IMPC, MGI, and CCDH, including sharing and editing mapping sets through Google Sheets.Source: Conclusions and Future Work · p. 20
  • The authors plan training materials, continued development of SSSOM and its Python toolkit, and expanded OxO support, while noting that a deliberately simple representation cannot capture every use case.They hope relevant communities will adopt the approach for more FAIR and reusable mapping sets.Source: Conclusions and Future Work · p. 20

Related papers

Loading 2112.07051v1…