Onto-what? Everything you (n)ever wanted to know about ontologies or were too embarrassed to ask

elixir-luxembourg.org Onto-what? Everything* you (n)ever wanted to know about ontologies or were too embarrassed to ask Danielle Welter Luxembourg National Data Service, ELIXIR Luxembourg *ok, maybe not everything - I only have 30 minutes So what CAN you tell me in half an hour? •Common ontology misconceptions - Too boring, too hard, too complex •The ontologist’s “toolbox” - Resources for (almost) every use case •Ontology use cases - Here’s some we made earlier Too boring, too hard, too complex Common ontology misconceptions Ontologies are only useful for data annotation! Why ontologies? ➔Reduce ambiguity ◆Shared understanding of complex concepts ◆Terms and relationships between them explicitly defined ◆Capture synonyms and cross-reference to other ontologies under one identifier ➔Human + machine readable ➔Integration, sharing, reproducibility How do we use ontologies? Data visualisation Data analysis Data integration Smarter searching Semantic models Ontologies are essential for… ➔Smarter searching: Improve search accuracy in biomedical databases by including synonyms, parent, child and sibling concepts in addition to classic lexical matching. ➔Data integration: Combine datasets from different sources using a common framework. ➔Text mining: Extract structured knowledge from unstructured biomedical literature. ➔Knowledge discovery: Reveal hidden patterns and relationships in biomedical data by leveraging the power of reasoning and inferred knowledge. Vocabularies, taxonomies, glossaries, ontologies… they’re all basically the same! Ontologies, taxonomies, vocabularies - same difference, right?! Ontologies are much too complex to be useful to anyone but experts! Are ontologies really too complex? ➔Why do ontologies mix user-understandable domain concepts with weird abstract terms! ◆ What even is a “specifically dependent continuant”?! ➔What to do when the same or similar terms appear in many ontologies? Upper-level ontologies - a common framework ➔Upper-level ontologies such as BFO or COB act as a common framework or “umbrella” under which independent ontologies can be linked OBI (Ontology for Biomedical Investigations) ENVO (Environment Ontology) Upper-level ontologies - BFO vs COB Basic Formal Ontology (BFO) ➔Domain-agnostic formal representation of upper-level concepts for interoperable domain ontologies Core Ontology for Biology and Biomedicine (COB) ➔Based on BFO but excludes some of the more confusing complexities So many ontologies to choose from… … how do I know which one is right? Single source of truth? Context is everything! Common ontology pitfalls ➔Improper reuse ◆Using ontology concepts in a context they weren’t intended for ➔Lack of reuse ◆I’ve got 99 terms with matching ontology annotations but this one concept ain’t one of them… Oh well, better build a new ontology! ➔Scope creep ◆Adding concepts to an ontology that lie outside its original scope ➔Not respecting constraints ◆Most ontology relations can only be used to connect specific types of concepts - a “domain” (source) and “range” (target) ◆Using them between incorrect types can badly break things What if no ontology perfectly fits my needs? Do you really need to build a new ontology? Why not get in touch with ontology developers and request new terms? Can I just build me own? Resources for (almost) every use case The ontologist’s “toolbox” Getting started - information, help & training ELIXIR RDM guidelines https://faircookbook.elixir-europe.org/ https://rdmkit.elixir-europe.org/ https://elixir-europe.org/what-we-offer/ guidelines/data-management https://tess.elixir-europe.org/ ●https://oboacademy.github.io/obook/ ●https://obofoundry.org/resources Example: The Open Targets Platform The Open Targets Platform (https://platform.opentargets.org/) integrates data from publicly available sources to support systematic identification and prioritisation of potential therapeutic drug targets. Extract of Figure 1 (part A-C) from Nucleic Acids Res, Volume 49, Issue D1, 8 January 2021, Pages D1302–D1310, https://doi.org/10.1093/nar/gkaa1027 Examples from the FAIR Cookbook •Creating a metadata profile for clinical trial protocols •https://w3id.org/faircookbook/FCB084 •Building a community compliant metadata profile - The Covid19 sample profile use case •https://w3id.org/faircookbook/FCB028 •Readying IMI Oncotrack - clinical cohort datasets for deposition to EBI Biosamples •http://w3id.org/faircookbook/FCB044 Any questions? Thank you ELIXIR RDM Community ELIXIR Interoperability Platform OBO Foundry Community ELIXIR resources representatives With special thanks to Get in touch! [email protected] OWL, OBO, JSON, RDF - acronym salad! ➔OWL (Web Ontology Language) ◆(family of) Knowledge representation language for ontologies ◆Widely used in all domains ➔OBO (Open Biomedical Ontologies) ◆File format for a biology-oriented language for building ontologies ◆Originally developed to be more user-friendly than OWL Standard tools are available to convert between OBO and OWL but OWL more widely used in modern ontologies ➔JSON & RDF ◆RDF/XML is the default serialisation for OWL ontologies ◆JSON is an alternative way of serialising ontologies