A HUMAN DEVELOPMENTAL KNOWLEDGE BASE FOR DISEASE INFORMATICS

CJ Mungall, C Torniai, JBL Bard, GV Gkoutos, S Essaid, PN Schofield, PN Robinson, D Smedley, M

Westerfield, SE Lewis, MA Haendel

From development to disorders

Development informs our understanding of diseases and disorders Congenital anomalies, cancer, etc

Can we build an informatics resource that encodes developmental knowledge?

○ Find all disorders that affect structures derived from the neural crest

○ Dynamically classify diseases based on developmental origin

○ What disease phenotypes are similar in terms of their developmental origins

○ What genes are implicated in disorders affecting X, and are differentially expressed in precursors of X?

neuralcrest

genes

disorders

neuralcrestderived structures

meckel’scartilage

ossicle

Challenges

bioinformatics databases

clinical resources

e.g. EHRs

?? SNOMED

ICD9

OpenBiologicalOntologies

(OBO Library)

1. Databases do not speak a common semantic language

2. Developmental relationships in mammalian ontologies are incomplete

neural tube

spinalcord

??embryoadult

“chasm ofsemanticdespair”

Creating a developmental KB: Approach Ontologies

Select open orthogonal ontologies covering domain Create bridging ontologies to cross chasm of despair Curate high quality developmental graph using OWL

axiom Infrastructure

Translate omics resources to OWL Build knowledge base on linked data cloud in modular

fashion using small ontologies and import chains Discovery

Query using fast EL++ reasoners and semantic similarity engines

embryonicanatomy

post-natal to adultanatomy

zebrafish mouse human

ontogenetic and phylogenetic knowledge transfer

homology

deve

lopm

ent

NL Washington, MA Haendel, CJ Mungallet al. Linking Human Diseases to Animal Models using Ontology-based Phenotype Annotation. PLoS Biology 2009

spinalcord

neural tube

Strategy for anatomy ontologies

MA FMA

EHDAA2EMAPA

Uberon

CJ Mungall, C Torniai, GV Gkoutos, SE Lewis, MA Haendel.Uberon, an integrative multi-species anatomy ontology. (2012) Genome biology 13 (1), R5

SNOMED(anatomy subset)

NCIt(anatomy subset)

ZFA

filling holes in developmental graphs

•taxon: mouse•use: gene expression (MGI)•adult only•no develops from relationships

MA

•taxon: zebrafish•use: gene expression and phenotypes (ZFIN)•adult and embryonic•463 develops from relationships

ZFA

•taxon: mouse•use: gene expression (MGI)•embryonic only•no develops from relationships

EMAPA

•taxon: human•adult only•no develops from relationships

FMA

•taxon: human-centric•adult and embryonic•no develops from relationships

SNOMED

filling holes in developmental graphs

•taxon: mouse•use: gene expression (MGI)•adult only•no develops from relationships

MA

•taxon: zebrafish•use: gene expression and phenotypes (ZFIN)•adult and embryonic•463 develops from relationships

ZFA

•taxon: mouse•use: gene expression (MGI)•embryonic only•no develops from relationships

EMAPA•taxon: human•embryonic only (CS1-20)•2108 develops from relationships

• high precision

EHDAA2

•taxon: human•adult only•no develops from relationships

FMA

•taxon: metazoa•adult and embryonic•783 develops from relationships•75 developmental contribution relationships•20 developmental induction relationships

Uberon

developmental relationships in uberon uberon includes ~900 developmental relationships

curated from literature, expert input classes in uberon are applicable across multiple species

parathyroid : tetrapods meninges: vertebrates

Challenge: developmental relationships vary throughout evolution

Solution: use OWL2 General Class Inclusion (GCI) axioms to encode

phylogenetically variable relationships Example:

○ (parathyroid and ‘part of some Aves) SubClassOf ‘has developmental contribution from’ some ‘ventral pouch of arch 3+4’

○ (parathyroid and ‘part of some Mammal) SubClassOf ‘has developmental contribution from’ some ‘dorsal pouch of arch 3+4’

http://uberon.org

The developmental logic of EHDAA2

inferior parathyroid

mesenchyme epithelium

‘has part’ o ‘develops from’ ‘has developmental contribution from’

3rd arch

3rd arch mesenchymefrom neural

crest

dorsal 3rd arch pouch endoderm

neural crest

endoderm

RO http://purl.obolibrary.org/obo/ro.owl

• precise assignment of develops from relationships

• subdivide organs by tissue type• e.g.

• mesenchyme/epithelium• leaf nodes in partonomy have full developmental lineage

http://www.obofoundry.org/wiki/index.php/EHDAA2:Main_Page

inferior parathyroid (Uberon) [mammal]

equivalence axioms allow integrative DL queries

inferior parathyroid

gland(FMA)

inferior parathyroid gland

(SNOMED)

‘has developmental contribution from’ some ‘neural crest’

OWL-DL Query

http://purl.obolibrary.org/obo/uberon/bridge/collected-mammal.owl

equivalent to:‘inferior parathyroid[mammal]’

and ‘part of’ some ‘Homo sapiens’

inferior parathyroid

mesenchyme epithelium

3rd arch

3rd arch mesenchymefrom neural

crest

dorsal 3rd arch pouch endoderm

neural crest

endoderm

integrating disorders and phenotypes Different disorder and phenotype resources use

different ontologiesMPO (mouse - MGD)HPO (human – OMIM, Orphanet)PATO+ZFA (zebrafish)SNOMED disorders, findings (human – EHRs)

Provide integrative definitional axiomsE.g. ‘neural tube defect’ EquivalentTo some morphological

abnormalitypato and ‘inheres in’ some ‘neural tube’uberon

Current work:Extending with axioms connecting GO developmental

processes to anatomical structures

CJ Mungall, GV Gkoutos, C Smith, MA Haendel, SE Lewis, M AshburnerIntegrating phenotype ontologies across multiple species. (2010) Genome biology 11 (1), R2

OmEO: omics data OmEO : Omics Entities Ontology

http://purl.obolibrary.org/obo/omeo Class-based obo-compliant representations of genes,

variants, transcripts, proteins, families and their relationships○ Sources:

ENSEMBL, GO, PANTHER, PRO, MODs

Expression data http://bgee.unil.ch/bgee/bgee

Functional annotation http://geneontology.org biogrid

Phenotype data Phenotype-commons

○ OMIM HPO○ MGI MPO○ ZFIN ZFA+GO_PATO

owl owl owl

importer

Using the knowledge base Current: Access via OWL tool chain

Reasoner: Elk UI: Protégé 4 Querying and semantic similarity searching

○ http://owltools.googlecode.com

Core URL:○ http://purl.obolibrary.org/obo/omeo/devkb.owl

imports multiple other ontologies via owl:imports chain

Future: web access via rdf triplestores LAMHDI Many components are available via neurocommons and

http://ontobee.org Challenge:

○ EL++ reasoning required

Current applications: enhancing semantic similarity queries Computing semantic similarity between

phenotypes find candidate disease genes, drugs find contributions of individual genes to multi-gene

phenotypes find commonalities between diseases

Adding developmental knowledge enhances similarity matching results available soon in mousefinder

○ http://wwwdev.ebi.ac.uk/panda-srv/mousefinder/mousefinder.php

CK Chen, CJ Mungall, GV Gkoutos et al.MouseFinder: candidate disease genes from mouse phenotype data. Human Mutation 2012

R Hoehndorf P Schofield, GV Gkoutos. PhenoNET: a whole phenome approach to disease gene discovery. NAR 2011

Conclusions Using uberon and ontology bridging axioms we can start

crossing the ‘chasm of semantic despair’ EHDAA2 and Uberon provide developmental graphs for

humans and a variety of other species Can be used to enhance existing ontologies (e.g. FMA,

SNOMED) OMEO DevKB integrates multiple ontologies and omics

resources Allows for queries and analyses that were not previously

possible Availability

http://uberon.org http://www.obofoundry.org/wiki/index.php/EHDAA2:Main_Page http://purl.obolibrary.org/obo/omeo/devkb

Acknowledgments Anatomy Ontologies

Melissa Haendel Terry Hayamizu Terry Meehan Alexander Diehl David Hill Brian Hall

Analysis Damian Smedley Rob Hoehndorf

OWL Infrastructure Carlo Torniai Davis Soumi-Sutherland Heiko Dietze Seth Carbon Allen Xiang Oliver He Alan Ruttenberg

EHDAA2 Jonathan Bard

Phenotype Ontologies George Gkoutos Paul Schofield Sandra Doelken Peter Robinson

OBO Foundry Barry Smith Richard Scheuermann Michael Ashburner Suzanna Lewis