Phenotype Ontologies Traversing All The Organisms (POTATO) workshop aims to reconcile logical definitions across species

Nicolas Matentzoglu, James P. Balhoff, Susan M. Bello, Cornelius F. Boerkoel, Yvonne M. Bradford, Leigh Carmody, Laurel Cooper, Christian A Grove, Nomi L. Harris, Sebastian Köhler, Marie‐Angélique Laporte, Stanley L.f. Laulederkind, Raymond Lee, Gaston K. Mazandu, Julie A. McMurry, Chris Mungall, David Osumi-Sutherland, Clare Pilgrim, Kayli Rageth, Sofia Robb, Peter N. Robinson, Erik Segerdell, Anne Thessen, Nicole Vasilevsky, Xingmin A. Zhang, Melissa HaendelZenodo 2018 · DOI

TL;DR

This paper introduces the POTATO workshop series, a collaborative effort to harmonize logical definitions of phenotypes across different species. By leveraging pattern-based ontology development, POTATO aims to overcome the challenges of semantic interoperability, enabling more accurate cross-species comparisons and data integration in biological research.

What Problem Does This Paper Address?

Biological research increasingly relies on phenotype ontologies to describe and classify observable characteristics of organisms. These ontologies use formal, logical definitions (often in OWL) to enable automated reasoning and data integration. However, a lack of consistent standards and manual coordination has led to divergent definitions for conceptually similar phenotypes across different species-specific ontologies (e.g., Human Phenotype Ontology - HP, Mammalian Phenotype Ontology - MP). This divergence hinders cross-species comparisons, making it difficult to identify homologous phenotypes or integrate data from diverse sources. For instance, "myocardial necrosis" is defined differently in HP and MP, leading to incompatible logical classifications and low semantic similarity scores.

Methods

The paper highlights the emergence of pattern-based ontology development as a solution. Specifically, it focuses on Dead Simple OWL Design Patterns (DOSDP). DOSDP provides a generic framework for defining reusable templates (patterns) for logical axioms, which can then be applied to create specific ontology terms. These patterns are stored in YAML files, and their applications are represented in tab-separated value (TSV) files. The Ontology Development Kit (ODK) is mentioned as a toolkit that supports DOSDP-based ontology curation. The POTATO workshop itself served as a venue for training in these tools and for collaborative exercises in reconciling divergent definitions. Participants worked in groups to evaluate definitions, develop new or adjusted EQ (Entity-Quality) definitions, and identify or create suitable DOSDP patterns.

Datasets

The paper does not explicitly list datasets used for training or evaluation in the traditional machine learning sense. Instead, it discusses the use of existing phenotype ontologies as the "data" for reconciliation. These include:

  • Human Phenotype Ontology (HP)
  • Mammalian Phenotype Ontology (MP)
  • Zebrafish Phenotype Ontology (ZP)
  • C. elegans (WBPhenotype)
  • Xenopus (XPO)
  • Drosophila (DPO)
  • Planarian (PLANP)
  • Cellular Microscopy (CMPO)
  • Plant Trait Ontology (TO)

The paper also refers to the UPHENO repository as a source of community-curated patterns.

Key Insights and Ideas

  • The "Entity-Quality" (EQ) approach is powerful but insufficient alone: While the EQ framework has advanced interoperability, manual coordination of its application across ontologies is not scalable and leads to divergence.
  • Pattern-based development is key to scalability: DOSDP offers a structured and reusable way to define logical axioms, promoting consistency and reducing manual effort.
  • Community collaboration is essential: Reconciling definitions requires active participation from ontology curators and developers across different model organisms. The POTATO workshop fostered this collaboration.
  • Divergent definitions don't always mean different meanings: The workshop revealed that many logical incompatibilities stem from differing pattern choices rather than fundamentally different biological concepts.
  • Challenges remain beyond definition reconciliation: Issues with the PATO (Phenotype and Trait Ontology) and UBERON (anatomy ontology), as well as capturing causality and handling phenotypic profiles, are significant hurdles.

Key Results

The POTATO workshop, attended by 40 participants, achieved several outcomes:

  • Training and Tool Introduction: Participants were trained in DOSDP-based ontology curation and introduced to relevant tools like ODK.
  • Collaborative Reconciliation: Six groups discussed and documented around 20 reconciliation problems, proposing or adjusting 5 concrete patterns. The example of 'unilateral deafness' was discussed by all groups, leading to comparable pattern proposals.
  • Community Engagement: A significant portion of attendees (half) committed to the ongoing reconciliation effort, with over 15 willing to dedicate an hour or more per week.
  • Identification of Challenges: The workshop highlighted specific issues with PATO (e.g., lack of textual definitions, inconsistent hierarchies) and UBERON (e.g., overly strict restrictions), as well as the need to better capture causality and handle phenotypic profiles.
  • Development of Tools: The Ontology Pattern Reconciliation Workbench was initiated to help identify and prioritize reconciliation candidates.

Limitations and Future Work

  • Scalability of Reconciliation: While pattern-based approaches improve scalability, fully reconciling thousands of definitions across multiple species is a multi-year effort.
  • Objective Prioritization: Determining which concepts need reconciliation and quantifying the impact of these efforts remains a challenge.
  • Tool Support: Further development of tools to automate the identification of reconciliation candidates (e.g., using string-matching or semantic integration) is needed.
  • PATO and UBERON Issues: The paper explicitly mentions the need to re-design PATO and address logical assumptions in UBERON for multi-species settings.
  • Phenotypic Profiles and Extensions: The formal vetting of phenotypic profiles and incorporation of phenotype extensions (e.g., unilateral deafness) into the methodology are areas for future work.
  • Ongoing Effort: The paper concludes by stating plans for future POTATO workshops, focusing on pressing issues like PATO redesign and UBERON compatibility. An ongoing program for training and migrating ontologies to DOSDP is also established.

Summary updated

Abstract

Logical definitions, in particular those following the Entity-Quality approach, are increasingly used to drive automated classification of phenotypes and integrate phenotypes across species semantically. Over the years, the lack of consistent and widespread use of common standards resulted in conceptually equivalent or similar phenotypes with logically divergent definitions. Recently, pattern-based approaches such as Dead Simple Ontology Design Patterns (DOSDP) have emerged as a way to reconcile such divergent definitions on a large scale. To facilitate the widespread adoption of DOSDP-based phenotype ontology development and to organise the reconciliation process involving key model organism communities, we have established the Phenotype Ontologies Traversing All The Organisms (POTATO) Workshop Series. These workshops are part of the NIH-funded “Forums for Integrative Phenomics” conference award, which provide a venue for phenotype ontology developers and phenomics informaticians to align their goals and technical approaches. The first POTATO workshop, in August 2018, provided training in cutting-edge tools and techniques for developing and releasing ontologies using design patterns and automated imports. This workshop also featured collaborative exercises in reconciling divergent logical definitions drawn from major phenotype ontologies that resulted in a set of general recommendations for how to improve phenotype ontologies and their development as well as a set of reconciled definitions. As a result of the workshop, many of the attendees have joined a phenotype ontology reconciliation effort, aimed at developing and implementing common design patterns across a range of phenotype ontologies. We thus plan to hold additional POTATO workshops.
BibTeX
@inproceedings{matentzoglu2018phenotype,
  author = {Nicolas Matentzoglu and James P. Balhoff and Susan M. Bello and Cornelius F. Boerkoel and Yvonne M. Bradford and Leigh Carmody and Laurel Cooper and Christian A Grove and Nomi L. Harris and Sebastian Köhler and Marie‐Angélique Laporte and Stanley L.F. Laulederkind and Raymond Lee and Gaston K. Mazandu and Julie A. McMurry and Chris Mungall and David Osumi-Sutherland and Clare Pilgrim and Kayli Rageth and Sofia Robb and Peter N. Robinson and Erik Segerdell and Anne Thessen and Nicole Vasilevsky and Xingmin A. Zhang and Melissa Haendel},
  title = {{Phenotype Ontologies Traversing All The Organisms (POTATO) workshop aims to reconcile logical definitions across species}},
  booktitle = {Zenodo},
  year = {2018},
  doi = {10.5281/zenodo.2382757},
}
Figure