Enhancing the Unified Phenotype Ontology to Support Cross-Species Phenotype Interoperability

4th uPheno Workshop, Biocuration, 5th April 2025

​

Organizers: Susan Bello, Arwa Ibrahim

Welcome to the 4th Unified Phenotype Ontology workshop!

• Welcome
• Introduction to uPheno (Nico Matentzoglu)
• uPheno Patterns - What are they and how do you use them (Sue Bello)
• Introduction of Challenges for the Workshop
• Making the Upper Level Structure Work
• Identify Structure and Integration Issues
• Breakout Groups to Work on the Challenges
• Wrap up of Breakout Group Results

uPheno in 2025 - Overview and Vision

An extreme simplification of the phenotype data workflow

Integration/Use

Observation

Coding/Curation

Clinician

Patient

Documentation

Clinical coder

EHR

Researcher

Specimen

Documentation

Biocurator

Scientific database

Code system

Ontology or vocabulary

?

More: https://obophenotype.github.io/upheno/reference/phenotype-data/#shape

Inconsistent language within the community

Source: Mammalian Phenotype Ontology

https://www.informatics.jax.org/vocab/mp_ontology/MP:0000274

macrocardia

enlarged heart

Image source: https://commons.wikimedia.org/wiki/File:Human_Heart.png

More inconsistent language across communities

Mouse community

Zebrafish community

MP:0000274

enlarged heart

ZFA:0000114, heart

PATO:0000586, increased size

Image source: https://commons.wikimedia.org/wiki/File:Human_Heart.png

Even more inconsistencies across the translational divide

Cardiomegaly

…

Mouse domain

Clinical domain

Zebrafish domain

Image source: https://commons.wikimedia.org/wiki/File:Human_Heart.png

Despite ontologies the inconsistent terminology problem got worse

ZFA/ZP

MP

DPO

WPO

HP

OMIA

VT

FYPO

APO

SNOMED

…

WB

PB

FB

OMIA

MGI

RGD

ZFIN

SGD

IMPC

OMIM

…ICD10

QTLdb

HPOA

EHR

XPO

XenBase

Phenotype core concepts 101

Characteristic

“Bearer”

Trait / biological attribute

phenotypic effect / phenotype

disease != phenotype

measurement != trait

https://obophenotype.github.io/upheno/reference/core-concepts/

Not just did ontologies proliferate, but also curation practices

• Different levels of standardisation
• Quantitative vs qualitative
• Post-coordinated vs pre-coordinated

More: https://obophenotype.github.io/upheno/reference/phenotype-data/#shape

An extreme simplification of the phenotype data workflow

Integration

Observation

Coding/Curation

Clinician

Patient

Documentation

Clinical coder

EHR

Researcher

Specimen

Documentation

Biocurator

Scientific database

Code system

Ontology or vocabulary

?

An extreme simplification of the phenotype data workflow

Integration

Observation

Coding/Curation

Clinician

Patient

Documentation

Clinical coder

EHR

Researcher

Specimen

Documentation

Biocurator

Scientific database

The Unified Phenotype Ontology

Code system

Ontology or vocabulary

The Unified Phenotype Ontology (uPheno)

• Integrates species specific phenotype ontologies using ontology design patterns
• Currently focussed on vertebrates, but some integration for yeast and slime molds
• Provides a computational framework for comparing phenotypes across species

The triumph of the Entity-Quality (EQ) pattern (Washington, 2009)

Quality

Entity

14

Hypolysinemia

=

Human phenotype

decreased amount

lysine

part of

blood

decreased circulating lysine level

=

Mouse phenotype

https://obophenotype.github.io/upheno/reference/data-integration/

uPheno is an ontology… and more

• Post-composed annotations are easily be turned into fully classified pre-composed ontologies
• ​

PMID:35317743

abnormal

Modifier

decreased amount

lysine

part of

Characteris.

Entity

blood

“decreased level of lysine in blood” (UPHENO:0034327)

​

Same for ZP..

.. but most importantly, it is a community that facilitates collective growth and development of shared best practices

We need a pattern for an abnormal biological process in a location.

More than 5 community members have rated this pattern as finished.

What is the best way to share the mappings?

How can I build a pattern driven ontology?

How can we support international phenotyping?

Participating ontologies

The three core use cases

1. Group phenotype data across species
2. Map between taxon-specific phenotypes and traits
3. Enable phenotypic profile comparisons across species (e.g. for diagnostics)

A

C

B

Where are we now, where are we going?

• uPheno has been deployed on OLS and in other applications for browsing and retrieval
• uPheno identifiers are finally stable and usable
• Publication out: PMID:39345458
• In progress:
• The hierarchy is still a bit wonky in places (today!)
• Some auto-generated classes are not very useful
• The integration with species specific phenotype ontologies is not as good as it could be
• (decide what to do about the scope)

uPheno Patterns - �What are they and how do you use them

Useful Links

• uPheno GitHub - https://github.com/obophenotype/upheno
• Please add issues here
• uPheno-dev GitHub - https://github.com/obophenotype/upheno-dev
• This has the latest release
• OLS uPheno browser - https://www.ebi.ac.uk/ols4/ontologies/upheno
• Monarch browser (re-ordered uPheno version) - https://ols.monarchinitiative.org/ontologies/upheno_reordered
• Phenotype ontologies Slack - https://join.slack.com/t/phenotype-ontologies/shared_invite/zt-31bgamfz8-P~T5AIZMe4pH64OnwMd5cw
• Protege - https://protege.stanford.edu/

Workshop channel in the ISB slack

#biocuration2025_upheno_workshop

uPheno Patterns

What are they and how do you use them

• Logical definitions (also called EQs)
• Machine readable definitions of phenotype terms built up using terms from other ontologies and connected by OBO Relations Ontology (RO) terms
• Example: enlarged heart (MP:0000274)

uPheno Patterns

• Variability in EQs makes it harder to integrate terms across ontologies
• Different developers may pick different PATO, anatomy, GO terms when creating an EQ
• Manually updating EQs, when decisions are made to change the way an EQ is constructed, is time consuming and error prone
• A shared pattern lets us
• Keep EQs consistent within and across ontologies
• Updates may be done using ROBOT either on demand or as part of an ontology build pipeline like those in ODK
• Updating the pattern file lets you update all terms using that pattern
• Pattern directory - https://github.com/obophenotype/upheno/tree/master/src/patterns/dosdp-patterns#pattern-directory

Useful background: Osumi-Sutherland, D., Courtot, M., Balhoff, J. et al. Dead simple OWL design patterns. J Biomed Semant 8, 18 (2017). https://doi.org/10.1186/s13326-017-0126-0

‘Characteristic of’ vs ‘Characteristic of part of’

• Prolonged discussions of when to use each of these relations
• ‘Characteristic of part of’
• Using this infers child terms down the tree(s) of the other ontology
• Example ‘morphology (PATO) and characteristic of part of (RO) some heart (UBERON)’
• Will include as child terms abnormal morphology of
• Cardiac valve (anatomy)
• Cardiomyocyte (Cell ontology)
• Mitochondrion located in Cardiomyocyte (GO cellular component combined with anatomy)
• We have agreed that use of ‘Characteristic of part of’ should be restricted to
• Morphology
• Physiology

Workshop Challenges

Workshop Challenges

1. Improve the upper level structure of the ontology

​

• Identify integration and organization issues

​

Challenge 1: Improve the upper level structure of the ontology

• These terms are the first 2-3 layers in the ontology

​

• Want these to help users browsing the ontology to intuitively find the correct branch for their terms of interest

​

• Often used for grouping annotations for display

Upper Level Structure

Upper Level Structure

Preferred Root

High level/header/grouping terms

Upper Level Structure

Preferred Root

1st level of division

2nd level of division

Single species example

Source: MGI

Single species example

Source: Monarch Initiative

Multi-species example

Source: MGI - HMDC tool

Multi-species example (not phenotype)

Source: Alliance of Genome Resources

Possible Approach

• Use specific subset tags to create multiple high level structures
• Allows customized sets of high level terms to coexist to serve multiple different use cases
• Users would need to agree on a set of terms and submit a GitHub ticket to get these added
• To work the terms need to be present and with the correct set of descendants
• We still want to avoid having redundant high level terms and we want terms to have the correct relationships to each other

Challenge 1 Questions

• In the re-ordered header version are there redundant top level terms?

​

• In the re-ordered header version are there missing top level terms?

​

Example

Why are all these absent terms directly under cell phenotype and not grouped in more useful ways?

Example

Why isn’t vocal organ phenotype a child of craniofacial/craniocervical phenotype?

Challenge 2: Identify integration and organization issues

• The various species specific ontologies have been integrated using the patterns and lexical mappings
• Not all terms have logical definitions, some complex phenotypes may be very hard/impossible to define logically
• Integration relies on anatomy integration in Uberon
• This integration is not perfect and we need help identifying areas where errors have occurred

​

Incorrect parent-child relationship example

Failure to group similar terms example 1

• EQs both use GO ‘ossification’ but have differences in how they are constructed
• Ideally these would be grouped under a uPheno term for ‘abnormal bone ossification’

Failure to group similar terms example 2

• Both terms use the same EQ
• Failure seems to be that the integration does not integrate abnormalities in a GO process that are part of the system

Break Out Sessions

• The main goal today is to identify as many problems as possible with uPheno.. and brainstorm ideas for solutions
• If you dare, create issues and add them to our priority backlog for 2025: https://github.com/orgs/obophenotype/projects/24/views/1
• Every breakout group should prepare a short report to be presented at 4:30 to the whole group
• Please call out any issues you see with uPheno, independent of your respective tasks

​

Please Add Notes to Shared Docs and Create Issues on the uPheno GitHub Repo

Summary From Breakouts

• Impossible to browse as is (virtual and in-person), can’t find your way through the maze, can’t find the label you want
• Need better search support
• Need really better synonyms
• Need better organization search results
• Classification philosophy in plant world - upper level is designed to meet use cases
• uPheno currently built using structures of other ontologies
• Can we get measures of how closely related some pairs of terms are

Keeping the Momentum Going

• Monthly editors meeting - contact Sarah (sarah@tislab.org) to receive an invitation to join
• Join the Phenotype-Ontologies Slack to keep in contact https://join.slack.com/t/phenotype-ontologies/shared_invite/zt-31bgamfz8-P~T5AIZMe4pH64OnwMd5cw
• Do you have a phenotype ontology you would like to see integrated into uPheno? Let us know on Slack!
• Do you have a use case you’d like to present? Let us know on Slack!