Modeling exposure events and adverse outcome pathways using ontologies
Download as PPTX, PDF0 likes229 views
This document discusses using ontologies to model exposure events, adverse outcome pathways, and phenotypes in order to support predictive toxicology. It describes existing ontologies like the Environment, Conditions, and Treatments Ontology (ECTO) and Gene Ontology Causal Activity Models (GO-CAMs) that can be used to represent exposure mechanisms and adverse outcomes. The document also presents challenges for developing an open predictive toxicology framework that leverages ontologies and linked data to make toxicology data more findable, accessible, interoperable, and reusable.
Modeling exposure events and adverse outcome pathways using ontologies
- 1. Modeling exposure events and adverse outcome pathways using ontologies Chris Mungall cjmungall@lbl.gov @chrismungall OpenTox USA 2018
- 2. Structure of talk Exposure Event Mechanism Adverse Outcome 12 3 Start here ontologies ontologies ontologies
- 3. Finding phenotypic data is hard Genetic or toxicogenomic database Search on: Hadziselimovic syndrome Results: 0
- 4. Ontologies and knowledge graphs: For FAIRness and predictive modeling Why ontologies? • Standard descriptors for annotating data • A shared model of the world to enable reasoning • Can be connected together to form Knowledge Graphs http://obofoundry.org/ontology/mondo.html
- 5. Ontologies: which ones? • Linked Open Vocabularies (LOV): – 650 vocabularies • BioPortal: – 716 ontologies – ~ 10m terms • Significant overlap in content • ‘Cancer’ found in 75 ontologies – Mappings of variable quality Kocbek PeerJ 2017 http://lov.okfn.org
- 6. Open Biological Ontologies (OBO) http://obofoundry.org 1. Well-integrated Modular ontologies 2. Provide technical and sociotechnological framework for cooperation 4. Enable feedback from community 3. Provide tools, best practices and infrastructure for forging new ontologies @obofoundry Workshop: how to build an OBO ontology ICBO August 2018 http://icbo2018.cgrb.oregonstate.edu/node/19
- 7. sup por ts sup por ts supports Chemistry, Genomes (CHEBI, SO) supports Cell Biology (GO) supports Development, Physiology (GO) supports supports supports Basic Descriptors (PATO) supports Cell Types (CL) supports Gross Anatomy (UBERON) supports supports supports supports Diseases (MONDO) supports supports supports Experiments (OBI) supports supports Environments (ENVO) supports Exposure (ECTO, ExO) sup por ts sup por ts sup por ts sup por ts Phenotypic features, symptoms (HPO, MPO) Part of the OBO ontology landscape - modular coupling - building block Diseases (MONDO)Adverse Events {
- 8. Palmoplantar hyperkeratosis (HPO) Hand (UBERON) Stratum corneum epithelial cell (CL) Hand (UBERON) Abnormally increased (PATO) Keratinization (GO) Köhler et al. (2014) F1000Research 2:30 Haendel et al. (2014) Modular term definitions offer many benefits
- 9. Ulcerated Paws (MP) Palmoplantar hyperkeratosis (HPO) Hand (UBERON) Stratum corneum epithelial cell (CL) Autopod (UBERON) Abnormally increased (PATO) Keratinization (GO) Köhler et al. (2014) F1000Research 2:30 Haendel et al. (2014) Modular term definitions offer many benefits
- 10. Ontology-based phenotype profile search www.owlsim.org Haendel, M., et al (2017) Drug Discovery Today: Disease Models. http://doi.org/10.1016/j.ddmod.2017.07.004
- 12. Smedley et al AJHG 2016 10.1016/j.ajhg.2016.07.005
- 13. KDM5B-related intellectual disability • Developmental delay • Multiple medical problems • Sees >5 hospital specialist services • Seen in two genetic centres • No cause known despite extensive testing • Now 4 years old • De novo truncation in KDM5B found – newly recognised disease gene • Exomiser ranked 2nd based on Protein-protein interaction to phenotype matched gene • Ends 4 year diagnostic odyssey • Informs parents on risk of recurrence in another child (very low) Georgia In Genomics England 100K Genomes, of first 1936 diagnosed patients, 82% are in the top 5 Exomiser hits across a range of rare diseases and family structures
- 15. We have a precise machine-readable language for describing some environmental exposures = CCOC(=O)CC(SP(=S)(OC)OC)C(=O)OCC CHEBI:6651
- 16. But others are harder to define Image: Zol87 CC by/nc
- 17. ECTO: Environment, Conditions and Treatments Ontology • Datamodel • Derived from ExO – Exposure Event – Stressor • Chemical • Macroscopic • Social – Receptor – Route • E.g. inhalation exhaustion https://github.com/EnvironmentOntology/environmental-exposure-ontology/ • Primary use cases – Gene-environment- phenotype studies – Integrating multi- species data, e.g. zebrafish
- 18. 2300 terms
- 19. 2300 terms
- 20. ECTO Construction ECTO ECTO CHEBI (chemical) ENVO environmental NPO nanoparticle SDGIO social UBERON Anatomy, routes PCI (populations) Osumi-Sutherland, D.,. (2017). Journal of Biomedical Semantics, 8(1), 18. http://doi.org/10.1186/s13326-017-0126-0 https://github.com/INCATools/ontology-starter-kit/
- 21. Uberon: bridging semantics for anatomy Mungall et al. (2012). Genome Biology, 13(1), R5. doi:10.1186/gb-2012-13-1-r5 JBMS 5:21 doi:10.1186/2041-1480-5- 21
- 22. Environment Ontology • Building block in construction of ECTO – Created originally for annotation of metagenomic samples • Contains classifications of – Environmental complex materials • Coal, oil, smoke, aerosols, etc – Environmental processes • Climate change, deforestation, … Buttigieg, P. L., (2016). Journal of Biomedical Semantics doi.org/10.1186/s13326-016-0097-6
- 23. Connecting ECTO to diseases and phenotypes • Current MONDO links; e.g. – Aflatoxin-related hepatocellular carcinoma – Asbestosis – Photosensitive disease • HPO annotations – Currently focused on monogenic diseases – Plans for disease-environment-phenotype
- 25. Modeling AOPs Image credit: Vinken, Toxicology, 2013 10.1016/j.tox.2013.08.011
- 28. Gene Ontology Annotation • Historically, functional annotation has used a ‘bag of terms’ model BFA1 S000003814 spindle pole GO:0000922 GTPase inhibitor activity GO:0005095 Exit from mitosis GO:0010458 GTPase activity GO:0003924 TEM1 S000004529 … …
- 29. A data model for causal ontology annotations: “GO-CAMs” Activity GO:nnnnnnn What: <gene product or complex> Where: GO/CL/Uberon Activity GO:nnnnnnn What: <gene product or complex> Where: GO/CL/Uberon Relationship RO:nnnnnnn Evidence: ECO, SEPIO Source: PMID, ORCID, ... GO-Causal Activity Model
- 30. Exit from mitosis GO:0010458 Example GTPase activity GO:0003924 What: TEM1 S000004529 Where: spindle pole GO:0000922 GTPase inhibitor activity GO:0005095 What: BFA1 S000003814 Where: spindle pole GO:0000922
- 31. Gene Ontology Causal Activity Models
- 32. http://noctua.berkeleybop.org/ Collaborative Editing! RDF/OWL Semantic Representation -Reasoning -Linked data Gene sets Building causal models of biology using ontologies
- 34. Mapping CTD events to GO-CAMs
- 36. Summary • Ontologies are necessary for FAIRness, but also predictive modeling and reasoning • Follow modular design patterns – Allows re-use of existing ontologies – Wider integration – Particularly for exposure ontologies • AOPs can be represented using GO-CAM model
- 37. Acknowledgments • Monarch – Kent Shefchek – Deepak Unni – Hyeongsik Kim – Nomi Harris – Tom Conlin – Tim Putman – Melissa Haendel – Peter Robinson – Tudor Groza – Damian Smedley – Sebastian Koehler – Jules Jacobsen • GO Causal Activity Models – Seth Carbon – Jim Balhoff – Tremayne Mushayama – Laurent-Phillippe Albou – Kimberley Van Auken – David Hill – Peter d’Eustachio – Huaiyu Mi – Dustin Ebert – David Osumi-Sutherland – Suzanna Lewis – Paul Thomas – GO Curators • OBO Services – James Overton – Rebecca Tauber – Suzanna Lewis – Eric Douglass – Randi Vita – Bjoern Peters • Environment/ECTO – Nicole Vasilevsky – Pier Luigi Buttigieg – Dan Keith
- 38. Challenges for an OpenTox framework • FAIR: – Making data • Findable • Accessible • Interoperable • Reusable • Leveraging curated knowledge to enhance predictive modeling • Diseases, pathways, genes, phenotypes, chemicals, …
- 39. Vision for a Predictive Toxicology Framework Toxicology Knowledge Commons Toxicology Data Commons Image credit: humantoxicologyproject.org Common Commons