The Semantic Web and Chemistry: Challenges, Opportunities, and why having a Canon of Chemical Ontologies can help

The Semantic Web and Chemistry Challenges, Opportunities, and why having a Canon of Chemical Ontologies can help 4th Ontologies4Chem Workshop 2025 November 11th2025 , Limburg an der Lahn (Hesse, Germany) Dr Samantha Pearman-Kanza (PSDI, University of Southampton) & Philip Strömert (NFDI4chem, TIB)  Outline ●History of the Semantic Web (Inception, Perceptions & Misconceptions)  ●The Semantic Web in Chemistry  ●Barriers, Challenges & Recommendations  ●Making the Semantic Web work for Chemistry NOW   ●Goals of this workshop Inception, Preconceptions & Misconceptions of the Semantic Web https://www.wi-consortium.org/wicweb/pdf/wi-hendler.pdf Our story actually begins 31 years ago… 1994 – First World Wide Web Conference – Tim Berners Lee shares his vision of the Semantic Web Debunking Common Misconceptions Berners-Lee, T. and Hendler, J., 2001. Publishing on the semantic web. Nature, 410(6832), pp.1023-1024.  Berners-Lee, T., Hendler, J. and Lassila, O., 2001. The semantic web. Scientific american, 284(5), pp.34-43. The Semantic Web - Critiques & Opinions •2014 – “Going through all the effort of putting semantic markup with no guarantee of a payoff for yourself was a stupid idea.” https://bibwild.wordpress.com/2014/10/28/is-the-semantic-web-still-a-thing/  •2016 - “Semantics is hard to understand”  https://www.linkedin.com/pulse/why-semantic-web-has-failed-kurt-cagle/   •2017 - “The Semantic Web was a great idea and still is. But I’m not seeing a future for it as it is. It needs to evolve and integrate it’s ideas with artificial intelligence” https://hackernoon.com/semantic-web-is-dead-long-live-the-ai-2a5ea0cf6423   •2022 – “in some ways, the Semantic Web was on life-support since its inception, and it continued to survive only with the medical intervention of academic departments who had no need to produce useable software or solve serious industry needs.” https://terminusdb.com/blog/the-semantic-web-is-dead/   •2025 – “Traditionally, you couldn’t say the word “ontology” in tech circles without getting a side-eye. Now? Everyone’s suddenly an ontology expert. And honestly… I’m here for it.” Juan Sequeda - https://www.linkedin.com/feed/update/urn:li:activity:7389287692645253120   But what is the Semantic Web Really? •The Web of Linked Data  •A way to bring context and meaning to data  •A set of common standards for data representation, integration, and search  •Used to create knowledge graphs and metadata https://www.sciencedirect.com/science/article/pii/B978012801238311520X?via%3Dihub Meanwhile in the Chemical Sciences Chemical Markup Language Introduction  •Chemical Markup Language (CML) - an XML-based format designed to represent and exchange chemical information (molecules, reactions, and spectra) in a structured, interoperable and machine-readable way.   •Introduced in 1995 by Peter Murray Rust and Henry Rzepa  Murray-Rust, P., Rzepa, H.S. and Leach, C., 1995. CML-chemical markup language. Chicago:[sn]   •Based on years of underpinning work around common formats:  Murray-Rust, P. and Rzepa, H.S., 2011. CML: Evolution and design. Journal of cheminformatics, 3(1), p.44.   •Key Features •XML-based: Built on standard web technologies for compatibility and extensibility. •Semantic Richness: Preserves the meaning of chemical data for both humans and machines. •Modular Design: Supports various chemical domains including molecular structures, reactions, and computational chemistry. •Validation: Uses conventions and dictionaries to ensure data integrity. https://www.ch.ic.ac.uk/rzepa/watoc96/wa_5.html  Ontology Barriers & Challenges •There are many ontologies “competing” in the same space, and yet large gaps still remain in others e.g. reactions  •“Lets just make a new one” mentality → Ontology Fatigue •Creating and maintaining ontologies to accurately represent aspects of the chemical sciences e.g. chemical entities and their properties is complex, and requires domain knowledge  •Ontology projects frequently don’t consider the full data lifecycle •competency questions aren’t always enough •ROI can be unclear when implementing in applications •Ontologies aren’t always standardised, well maintained or FAIR •Harmonizing/Aligning ontologies across the chemical sciences is complex, time consuming, and technically challenging Image created using https://openart.ai/  Ontology Mitigations •Re-use and extend where possible  •Utilise Ontology Lookup Services to identify potentially relevant ontologies  •Re-use doesn’t just mean ontologies - you can re-use design patterns as well e.g. OBO Foundry •Consider extending existing ontologies  •Modularise Ontologies •Break them down into smaller related modules for ease of use and improved re-use •Use Software Development Approaches •Ontology Development Kit & Git •Continuous Integration & Testing https://www.fosteropenscience.eu/content/cartoonreusable-data Examples of Ontology Best Practice OBO Foundry Data Barriers •Chemical Sciences data comes from disparate sources and is available in many formats, making it hard to standardise  •Converting ALL your data into knowledge graphs is a huge undertaking (and storage intensive)  •Some data doesn’t lend itself to being represented semantically e.g. numerical classifications   •CONVERTING POOR QUALITY DATA TO LINKED DATA DOESN’T NECESSARILY FIX IT https://imgflip.com/memegenerator/ Data Mitigations •You need to ask the right questions! •What is your use case for semantics? •Do you need a full knowledge graph?  •What actually needs to be represented semantically? •Do you need Semantic Data or Semantic Metadata?  •How can AI/LLMs be leveraged here? •Researchers are starting to consider how to automate the generation of semantic metadata  CC BY-ND 3.0 Dataedo - https://dataedo.com/cartoon/ Examples of Best Practice Noticeable shift towards creating Semantic Metadata rather than marking up entire datasets as RDF •Creating a Semantic Metadata wrapper •Using JSON-LD (either with DCAT  or schema.org) •Using RO-Crate with JSON-LD Focus on how to find the right dataset •Which molecules, analysis methods, reactions, experimental conditions where involved in the creation of a dataset  → e.g. nfdi-de.github.io/chem-dcat-ap  However, many challenges still remain Chemistry related terms are defined in multiple ontologies.  ➔Many were born 10+ years ago  ➔All are incomplete → always ➔Which ontologies should we use when? ➔Should every chemist need to know all about them? Ontology Alignment/Harmonization & Development is complex. ➔We are still working on many foundational issues ➔Ontology Fatigue is always just around the corner ➔But it is worth it! → now: FAIR data, later: better AI ➔We now have better tooling to make this work ➔Let’s push this as another form of scientific contribution Chemical Terms Defined in Multiple Ontologies •Do you know which ontology covers what and what not? •Do you know the ontological commitments entailed? •Do you know about its quality, maintenance status and current user base? Ontologies4Chem https://vevox.app/#/m/11 1232520 Session ID: 111-232-520 Chemical Terms Defined in Multiple Ontologies Wouldn’t a canon of chemical ontologies you can abide by help? Let’s create the Ontologies4Chem Canon CANON (Ontological, n.) [from Gk. κανών (kanon), meaning "rule"; redefined for digital science] 1. A Living, Foundational Hypertext defining the essential collection of ontologies for a given domain. 2. An essential collection of ontologies whose value derives not from static completeness but from the active, communal engagement of domain experts and ontologists. 3. Usage: To be constantly interpreted, critiqued, annotated, and formalized by the community to ensure its continuous relevance and utility for the evolving needs of digital science. Harmonization & Development Is Complex But we now have better tooling •Protégé •Maintained by the open science community •ROBOT •powerful command line tool & library •extract, merge, reason, convert, diff, … •Ontology Development Kit (ODK)  •allows us to develop/maintain in a standardized manner •depends on ROBOT and GitHub/GitLab We Now have Better Tooling Documentation is part of the tooling! oboacademy.github.io/obook •There is plenty of great training material already. •It will be better the more we use and improve it. •Also good for interested domain experts.  We Now have Better Tooling •Git (publishing on GitLab or GitHub) •transparent & automatized management •improves community curation •Ontology Look-Up Services •find, read & understand ontologies •MOD: shared publication standards •a single point of truth | first point of entry  Our access point for the  Ontologies4Chem Canon ●An intuitive way to work with the indexed ontologies ●Based on EBI’s Ontology Lookup Service (OLS) ●GitHub integration,  ●Commenting on ontologies or terms,  ●advanced search,  ●sharable custom ontology collections and term sets 35 Building the Canon in the NFDI4Chem TS 35 ●An intuitive way to work with the indexed ontologies ●Based on EBI’s Ontology Lookup Service (OLS) ●GitHub integration,  ●Commenting on ontologies or terms,  ●advanced search,  ●sharable custom ontology collections and term sets 36 Building the Canon in the NFDI4Chem TS 36 ●An intuitive way to work with the indexed ontologies ●Based on EBI’s Ontology Lookup Service (OLS) ●GitHub integration,  ●advanced search,  ●sharable custom ontology collections and term sets 37 Building the Canon in the NFDI4Chem TS 37 ●An intuitive way to explore chemical ontologies ●Based on EBI’s Ontology Lookup Service (OLS) plus ●GitHub integration,  ●advanced search,  ●sharable custom ontology collections and term sets, 38 Building the Canon in the NFDI4Chem TS 38 ●An intuitive way to work with the indexed ontologies ●Based on EBI’s Ontology Lookup Service (OLS) ●GitHub integration,  ●advanced search,  ●sharable custom ontology collections and term sets, ●commenting on ontologies or terms  ● 39 Building the Canon in the NFDI4Chem TS 39 Current Pain Points to Tackle Oldies but Goldies ●How and when to “revive” un-/under-maintained ontologies? ○FIX - Physico-chemical methods and properties ○REX - Physico-chemical process ○GC - Gainesville Core Ontology ○CHEMINF ●Let us try to ○add context ○files issues ○fix issues Time/Day Mon, Nov 10 Tue, Nov 11 Wed, Nov 12 Thu, Nov 13 08:00 - 09:00 Breakfast 09:00 - 10:00 Welcome & Opening Keynote Talks around the topic “Ontological Representation of Chemical Reactions” Talks around the Topic “Ontological Representation of Chemical Entities” 10:00 - 12:00 Talks from NFDI consortia, PSDI etc. 12:00 - 13:00 Lunch 13:00 - 16:00 Arrival & Registration Updates from Ontology Projects & Applications build upon ontologies Hacking Sessions Hacking Sessions Preparation for the Final Round 16:00 - 18:00 Hacking Sessions Ontology Curation & Tools Social Event Final Round (Reporting back on Hacking Session / Final Discussion) & Closing of the Workshop 19:30 Dinner Current Pain Points to Tackle Chemical Reactions  ●What can we represent already? ●What do we need to define where and how? ●How to use the OntoCape based ontologies? ○If you use it, please help us improve the documentation! Time/Day Mon, Nov 10 Tue, Nov 11 Wed, Nov 12 Thu, Nov 13 08:00 - 09:00 Breakfast 09:00 - 10:00 Welcome & Opening Keynote Talks around the topic “Ontological Representation of Chemical Reactions” Talks around the Topic “Ontological Representation of Chemical Entities” 10:00 - 12:00 Talks from NFDI consortia, PSDI etc. 12:00 - 13:00 Lunch 13:00 - 16:00 Arrival & Registration Updates from Ontology Projects & Applications build upon ontologies Hacking Sessions Hacking Sessions Preparation for the Final Round 16:00 - 18:00 Hacking Sessions Ontology Curation & Tools Social Event Final Round (Reporting back on Hacking Session / Final Discussion) & Closing of the Workshop 19:30 Dinner