Why Science? | Core Ontology for Biology and Biomedicine

Science learning becomes powerful when specialised vocabularies share a small, dependable semantic floor. The OBO Foundry describes the Core Ontology for Biology and Biomedicine, or COB, as a small ontology that brings together key terms from a wide range of OBO projects to improve interoperability and reuse. Its official page lists a main ontology, base module, native-ID module, external mapping product and an explicitly experimental demo product. The official repository says the plan is to keep COB small while ensuring one or more COB terms can serve as roots for OBO Library ontologies. COB can coordinate high-level meaning, but it does not replace domain definitions, evidence, releases or the scientific judgement needed to map specialised data. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science, O-Level Biology, computing and STEM education as a longform evidence-reading route; it is not medical, laboratory, admission or career advice.

Related eduKate reading: Relation Ontology; Information Artifact Ontology; Ontology for Biomedical Investigations; Gene Ontology; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.

Primary and official sources checked on 13 October 2026: OBO Foundry Core Ontology for Biology and Biomedicine overview, checked 13 October 2026; Official COB source repository; Canonical COB OWL release; Official COB project documentation; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Ontologies, definitions and release products can change, so preserve dated documentation, exact source records and file versions.

How a small core ontology supports a large scientific ecosystem

Scientific ontologies often specialise deeply: one may describe cells, another experiments and another gene functions. Shared upper terms help their classes meet without forcing every project to rebuild the same roots.

The OBO Foundry says COB deliberately stays small. Smallness is a design constraint: the core should coordinate concepts that many ontologies need while leaving domain detail with the communities that understand it.

The official repository describes COB terms as possible roots for OBO Library ontologies. A root supports organisation and reasoning, but it does not make every descendant concept equivalent across domains.

COB publishes multiple products. The base module contains COB-ID terms, while fuller products can include external OBO terms used in their axioms; product choice changes what a system imports.

The native module preserves native identifiers rather than rewiring them. Identifier strategy affects provenance, maintenance and how downstream systems recognise the authority for a term.

An external mapping product can describe connections without pretending the core owns every specialised class. Mappings need scope because exact equivalence is stronger than useful relatedness.

The official page marks the demo product as experimental. Experimental examples can teach and test design ideas, but production claims should name their status and avoid treating a demonstration as normative.

The hopeful value is coordination without erasure: projects gain a shared semantic floor while retaining the definitions, evidence and nuance that make specialised science useful.

Begin with competency questions. If two datasets must answer a cross-domain query, write the expected results and counterexamples before choosing a core term.

Keep imported and native identifiers distinguishable. A familiar label can conceal which project defines the class and which release supplied its axioms.

Module boundaries affect performance and reasoning. Importing a smaller core may simplify deployment, while a fuller module can provide connections needed for richer inference.

Logical consistency is necessary but not sufficient. A mapping can be formally satisfiable and still misrepresent the biology or study design.

Interoperability is tested by tasks, not by file loading alone. Queries, data exchange, validation and round-trip conversion reveal whether meaning survived.

Core terms should not become a dumping ground for unresolved local concepts. Ambiguous data may need an explicit provisional status rather than a broad class chosen for convenience.

For classroom learning, a shared concept map can show how cell, organism, process, information and experiment link while each science topic keeps its own vocabulary.

Reader-ready documentation names the selected COB module, external dependencies, mapping rules and known losses so another team can reproduce the integration.

Version drift can split an ecosystem when participating ontologies import different core releases. Dependency coordination and tested upgrade windows reduce silent semantic divergence.

Rewiring identifiers may simplify a graph while obscuring the source authority. Preserve attribution and migration history whenever identifiers are transformed.

Reasoner-generated classifications should be separated from curator assertions. Both may be useful, but they answer different provenance questions.

A crosswalk can be asymmetric: one specialised class may map broadly to a core term while the reverse mapping would return many scientifically different descendants.

Closed validation rules and open-world ontology reasoning solve different problems. Use the first to enforce required records and the second to derive declared logical consequences.

Metrics such as fewer unmapped terms or more shared roots are not proof of better science. Review whether the resulting queries preserve distinctions users actually need.

A reproducible ontology assembly freezes products, import catalogues, checksums, reasoner, mapping code, test fixtures and query outputs alongside human review notes.

COB succeeds when shared concepts make collaboration easier without becoming an excuse to compress every scientific disagreement into one high-level label.

Governance is part of interoperability. A shared core term may affect many downstream ontologies, validators and datasets, so proposed changes need visible discussion, versioning and impact analysis. The authority responsible for a specialised class may differ from the group maintaining its high-level parent. Record who can approve definition changes, how deprecations are communicated and which compatibility promises apply. This prevents a technical convenience from silently moving scientific authority away from the community that collects and interprets the evidence. It also gives adopters a route for reporting edge cases that the current core does not represent well.

Interoperability tests should include round trips. Export a specialised record through the COB-aligned exchange model, import it back and compare identifiers, relations, qualifiers and provenance with the source. A successful parse is only the first checkpoint. The restored record should answer the same scientific questions and should not acquire broader claims through unintended inheritance. Counterexamples are especially valuable: include terms that must remain distinct even though they share a core parent. These fixtures make semantic loss observable and convert vague confidence in a shared ontology into a repeatable quality-control process.

Core ontologies can reduce duplication, but aggressive consolidation may increase coupling. If every domain change depends on one central release, maintenance can slow and local innovation can become risky. Modular design offers a middle path: coordinate the concepts that genuinely need to be shared, publish stable interfaces and let domain ontologies evolve behind them. Teams should measure upgrade effort, failed mappings and query changes over time. A small core is successful not because all projects look identical, but because necessary connections remain intelligible, testable and recoverable when one participating ontology changes.

For learners, COB offers a useful systems lesson. Biology is not a flat vocabulary list; it is a network in which organisms have parts, processes occur, investigations produce information and evidence supports claims. A core ontology shows the common scaffolding, while specialised ontologies show why detail matters. Students can build a five-node concept map, label each relation and then compare it with a second topic. The goal is not to memorise ontology engineering terms. It is to practise moving between a big organising idea and the precise local concept needed to explain an observation honestly.

A four-stage evidence ladder for COB-backed interoperability

Stage	Question	Evidence	Responsible output
Scope	Which high-level concept must be shared?	Competency question and participating ontologies	A bounded interoperability need
Select	Which COB product and identifiers apply?	Base, full or native module with frozen release	A declared semantic core
Map	How do specialised terms connect to it?	Definitions, axioms, mappings and provenance	Traceable cross-ontology alignment
Test	Do expected queries work without lost meaning?	Fixtures, reasoner output and domain review	Validated interoperability