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