Expansion of the Human Phenotype Ontology (HPO) knowledge base and resources

Sebastian Köhler, Leigh Carmody, Nicole A. Vasilevsky, Julius O. B. Jacobsen, Daniel Danis, Jean-Philippe F. Gourdine, Michael A. Gargano, Nomi L. Harris, Nicolas Matentzoglu, Julie A. McMurry, David Osumi-Sutherland, Valentina Cipriani, James P. Balhoff, Tom Conlin, Hannah Blau, Gareth Baynam, Richard Palmer, Dylan Gratian, Hugh J. S. Dawkins, Michael Segal, Anna C. Jansen, Ahmed Muaz, Willie H. Chang, Jenna Bergerson, Stanley J. F. Laulederkind, Zafer Yüksel, Sergi Beltran, Alexandra F. Freeman, Panagiotis I. Sergouniotis, Daniel Durkin, Andrea L. Storm, Marc Hanauer, Michael Brudno, Susan M. Bello, Murat Sincan, Kayli Rageth, Matthew T. Wheeler, Renske Oegema, Halima Lourghi, Maria G. Della Rocca, Rachel Thompson, Francisco Castellanos, James Priest, Charlotte Cunningham-Rundles, Ayushi Hegde, Ruth C. Lovering, Catherine Hajek, Annie Olry, Luigi Notarangelo, Morgan Similuk, Xingmin Aaron Zhang, David Gómez-Andrés, Hanns Lochmüller, Hélène Dollfus, Sergio Rosenzweig, Shruti Marwaha, Ana Rath, Kathleen Sullivan, Cynthia L. Smith, Joshua D. Milner, Dorothée Leroux, Cornelius F. Boerkoel, Amy Klion, Melody C. Carter, Tudor Groza, Damian Smedley, Melissa A. Haendel, Chris Mungall, Peter N. RobinsonNucleic Acids Res 2019 · DOI

TL;DR

The Human Phenotype Ontology (HPO) is a crucial resource for standardizing the description of phenotypic abnormalities in rare diseases, aiding in diagnosis and research. This paper details significant expansions and improvements to the HPO, including new terms, enhanced logical definitions, better interoperability with other ontologies, and a redesigned website, making it more powerful and accessible for a growing user base.

What Problem Does This Paper Address?

Accurately diagnosing rare diseases is a monumental challenge due to their heterogeneity and the vast number of potential conditions. A key bottleneck is "deep phenotyping" – the detailed and standardized description of an individual's clinical abnormalities. Without a common language, comparing patient data, identifying patterns, and linking phenotypes to genetic causes is incredibly difficult. The Human Phenotype Ontology (HPO) was developed to address this by providing a standardized vocabulary for phenotypic abnormalities. However, as the field of rare disease research and diagnostics evolves, the HPO needs to continuously expand its knowledge base, improve its logical structure for computational use, and enhance its accessibility to researchers, clinicians, and even patients. This paper tackles the problem of keeping the HPO a cutting-edge, comprehensive, and user-friendly resource in the face of these evolving needs.

Methods

The authors describe several key methodological advancements and resource expansions:

  • Content Expansion: The HPO has been significantly expanded with new terms, particularly in areas like molecular phenotypes derived from metabolomics data.
  • Enhanced Logical Definitions: The paper highlights the use of pattern-based ontology development (e.g., Dead Simple OWL Design Patterns - DOSDP) to create and maintain compatible logical definitions across multiple phenotype ontologies (HPO, Mammalian Phenotype Ontology, etc.). This approach aims to reduce manual errors and improve consistency, enabling better cross-species comparisons and computational reasoning.
  • Interoperability: The HPO's logical interoperability with other ontologies like Uberon (anatomy) and Gene Ontology (function) is leveraged, particularly within tools like Exomiser for variant prioritization.
  • New Subontologies: The introduction of "Clinical modifier" and "Clinical course" subontologies allows for more nuanced and precise descriptions of disease features, including severity, positionality, triggers, and temporal patterns.
  • Redesigned Website: A new, single-page web application for the HPO website has been developed, focusing on improved responsiveness, intuitive browsing, auto-complete search, and term highlighting.
  • Layperson Synonyms: To improve patient engagement, "layperson" synonyms have been added to a significant portion of HPO terms, making them more understandable to individuals without medical expertise.
  • Updated Annotation File Format: The HPO annotation file format has been revised to accommodate new features like clinical modifiers and sex-specific disease characteristics.

Datasets

The paper does not explicitly list specific datasets used for training or evaluation in the traditional machine learning sense. Instead, it discusses the HPO's integration with and use of various existing data resources:

  • Metabolomics data: Used to identify new molecular phenotypes.
  • Human Metabolome Database (HMDB): A source for metabolomics data.
  • Clinical data from rare disease databases and projects: Such as the Undiagnosed Diseases Network (UDN), RD-Connect, and the 100,000 Genomes Project, which contribute to HPO annotations and disease definitions.
  • Literature: A primary source for curating disease annotations and identifying new terms.
  • Existing ontologies: Such as MONDO (Monarch Merged Disease Ontology), OMIM, Orphanet, Uberon, and Gene Ontology, which are integrated with HPO.

Key Insights and Ideas

  • The Power of Standardization: The HPO's success underscores the critical need for standardized vocabularies in complex biological and medical domains. It's not just about listing terms, but about creating a structured knowledge base that machines can understand and reason with.
  • Pattern-Based Ontology Development is Key: Manually maintaining complex ontologies is prone to errors. The adoption of pattern-based development (DOSDP) is a significant step towards scalable, consistent, and interoperable ontology creation, especially across different species.
  • Bridging the Gap Between Clinicians and Patients: The inclusion of "layperson" synonyms is a vital step towards patient empowerment and improved communication in rare disease care.
  • Interoperability is Not Optional: The HPO's utility is amplified by its seamless integration with other ontologies and tools (like Exomiser), demonstrating that isolated ontologies have limited impact.
  • Community Collaboration Drives Progress: The paper emphasizes the importance of workshops and community input in refining and expanding the HPO, highlighting a collaborative model for knowledge base development.

Key Results

  • Significant Content Growth: The HPO has been expanded with new terms, particularly in molecular phenotypes.
  • Improved Logical Definitions: The use of pattern-based development is enabling more robust and consistent logical definitions, facilitating computational analysis and cross-species comparisons.
  • Enhanced Usability: The redesigned HPO website offers a more intuitive and efficient user experience for browsing terms and exploring related data.
  • Increased Accessibility: The addition of layperson synonyms and ongoing translation efforts are making the HPO more accessible to a broader audience, including patients.
  • Integration with Major Initiatives: The HPO continues to be adopted by major rare disease initiatives and tools, solidifying its role as a de facto standard.

Limitations and Future Work

  • Coverage of Layperson Synonyms: While significant, layperson synonyms are not yet available for all HPO terms, and their diagnostic utility is still under evaluation.
  • Manual Curation Challenges: Despite pattern-based development, manual curation remains essential and can be error-prone.
  • Ongoing Development: The authors indicate that efforts are continuously underway to improve specific areas of the ontology, extend standardized disease descriptions, and integrate with external community data. Future work will likely focus on further expanding the knowledge base, refining logical definitions, and developing new computational tools and integrations. The HPO project actively welcomes interactions with the community to address these ongoing needs.

Summary updated

BibTeX
@article{kohler2019expansion,
  author = {Sebastian Köhler and Leigh Carmody and Nicole A. Vasilevsky and Julius O. B. Jacobsen and Daniel Danis and Jean-Philippe F. Gourdine and Michael A. Gargano and Nomi L. Harris and Nicolas Matentzoglu and Julie A. McMurry and David Osumi-Sutherland and Valentina Cipriani and James P. Balhoff and Tom Conlin and Hannah Blau and Gareth Baynam and Richard Palmer and Dylan Gratian and Hugh J. S. Dawkins and Michael Segal and Anna C. Jansen and Ahmed Muaz and Willie H. Chang and Jenna Bergerson and Stanley J. F. Laulederkind and Zafer Yüksel and Sergi Beltran and Alexandra F. Freeman and Panagiotis I. Sergouniotis and Daniel Durkin and Andrea L. Storm and Marc Hanauer and Michael Brudno and Susan M. Bello and Murat Sincan and Kayli Rageth and Matthew T. Wheeler and Renske Oegema and Halima Lourghi and Maria G. Della Rocca and Rachel Thompson and Francisco Castellanos and James Priest and Charlotte Cunningham-Rundles and Ayushi Hegde and Ruth C. Lovering and Catherine Hajek and Annie Olry and Luigi Notarangelo and Morgan Similuk and Xingmin Aaron Zhang and David Gómez-Andrés and Hanns Lochmüller and Hélène Dollfus and Sergio Rosenzweig and Shruti Marwaha and Ana Rath and Kathleen Sullivan and Cynthia L. Smith and Joshua D. Milner and Dorothée Leroux and Cornelius F. Boerkoel and Amy Klion and Melody C. Carter and Tudor Groza and Damian Smedley and Melissa A. Haendel and Chris Mungall and Peter N. Robinson},
  title = {{Expansion of the Human Phenotype Ontology (HPO) knowledge base and resources}},
  journal = {Nucleic Acids Res.},
  year = {2019},
  doi = {10.1093/nar/gky1105},
}
Figure