Integrating phenotype ontologies across multiple species

Integrating phenotype ontologies across multiple speciesCaltech 2009
Chris Mungall

Conservation of phenotypes

How do we compare phenotypes across species?
Multiple project or species specific phenotype ontologies
MP: mouse
HP: human
WP: worm
TO: plant
APO: fungi
Data is not comparable
How do we know they are built according to the same principles?

Anatomical and structural entities in OBO

MP
MP

is_a hierarchy
hand-crafted

PATO EQ Descriptions
PATO:
ontology of qualities
EQ Descriptions
An expression consisting of:
a PATO ID (Q)
A bearer entity (E)
Optional additional tags

Examples

Relational Qualities
Have a bearer
- have an additional dependency on some type of entity

EQ expressions can be written in obo format
Why?
Formal semantics
Edit within OBO-Edit as genus-differentia definitions
Use the OBO-Edit reasoner
Can also be translated to OWL

EQ expressions can be written in obo format

EQ vs dedicated phenotype ontology
Some groups pre-compose phenotype descriptions
terms generated in advance by ontology editor
ontology editors adds defs etc
dedicated phenotype ontology
often species-specific
Other groups post-compose phenotype descriptions
descriptions composed as needed by annotators
no dedicated phenotype ontology

Methodologies

Unifying descriptions
Equivalence Mappings
between
term
EQ expression

Methods
Use obol to generate first pass and updates
parses term name
assumes semi-controlled syntax
Manually edit obo xp file

Results

Advantages
Reuse PATO across ontologies
Make sure we’re saying the same thing
We can use the OBO-Edit reasoner to compute the is_a hierarchy
We can enhance queries within species.
E.g. find all genes that have a phenotype affecting sperm (CL:0000019 / WBbt:0006798)
We can compare across species

Reasoning

Reasoner Results: HP/MP

Comparing across species
Requirements:
Phenotype ontology mappings
Mappings between anatomy ontologies
Uberon
Multi-species anatomy ontology
No assumption of homology
Analogy is useful to; e.g. eyes
Scope: metazoans
Applicability outside vertebrates controversial
Better to focus on cells (CL)

Mapping AOs to Uberon

Phenoblast: Quantifying similarity
SimJ: 0.42
MaxIC: 13.4
SimJ: 0.17
MaxIC: 6.2
SimJ: 0.32
MaxIC: 12.1

Recovering
pathway
members
by phenotypic
similarity

Conclusions
Sharing ontologies is good
PATO / EQs can be used to help build ontologies as well as in annotation
Find mistakes
Automatically build DAG (reasoning)
Mapping a phenotype ontology to EQs gives you the best of both worlds
Control over the terms used in annotation
Comparing data across species
Where anatomical similarities permit..