Dental science · Formal ontology

Walk the ontology of a cavity

Dentistry has a formal, machine-readable definition of dental caries. It is rigorous, it is open, almost nobody has looked at it — and it stops at exactly the interesting place.

Joshua Grenier, DMD

Somewhere in a GitHub repository maintained by three people, dentistry has a formal definition of a cavity. Not a textbook description — a machine-readable one, with logical axioms, that a reasoner can compute over.

Most dentists have never seen it.

What this is

An ontology is not a dictionary. A dictionary tells you what words mean. An ontology asserts what exists and how those things connect — as a graph you can traverse and a machine can reason over.

Dentistry already has one formal vocabulary, and every practice uses it daily: CDT procedure codes. But a code records what was done. D2392 does not tell a computer what a cavity is, what causes one, or what it does to a tooth. It tells an insurer what to pay for.

The Oral Health and Disease Ontology — OHD — is the other thing. Built at the University at Buffalo School of Dental Medicine under NIDCR funding by Alan Ruttenberg, Bill Duncan and Titus Schleyer, it is the only artifact I know of that attempts to say, formally, what dental disease is.

It holds 1,776 classes. Roughly half are borrowed — anatomy from UBERON, chemistry from ChEBI, the disease framework from OGMS — all resting on Basic Formal Ontology as an upper layer. The rest is dental.

The three-way split

OHD does not have one thing called caries. It has three, and keeping them apart resolves a confusion clinicians live with daily.

Carious tooth lesionMaterial entity

A physical thing. OHD’s definition is unusually good: it covers not just demineralized tooth structure but the infectious organisms, the material they generate from the tooth, and any immune effectors responding to it. The biofilm is part of the lesion, not adjacent to it.

Dental cariesDisposition

A tendency the lesion bears, the way a glass is fragile before it breaks. Dispositions exist whether or not they are being exercised.

Bacterial tooth demineralizationProcess

The disposition actually realized. Acid actually dissolving mineral, right now.

Dentistry says “caries” for all three and cannot tell them apart. That is not pedantry. It is why arrested caries sounds like a contradiction and is not: the lesion persists, the process has stopped, and whether the disposition remains is the actual clinical question — the one that decides whether you restore or watch.

OHD has the term. Arrested carious lesion is defined as a lesion that resulted from some tooth decay but where the infected region is not currently decaying.

Where it stops

The ontology contains a class called dental caries disease course. It has a definition. It has nothing beneath it.

···

Across all 1,776 classes, terms containing stage, rate, progression or transition return three hits. OHD defines six relations of its own, and the two temporal ones — next visit and later visit — order appointments, not disease states. No relation anywhere says one pathological condition develops into another.

This is a boundary, not sloppiness. The formalism can say what things are, what they are part of, and what they are about. It cannot express a rate, a threshold, or a probability. You cannot write “CAL ≥ 5 mm” as a logical definition, and you cannot write “this lesion progresses to that one with probability p.”

We need to hook up the disease with the material basis, and possibly a representation of the disease course. However we don’t have the ‘has material basis’ relation yet, and it’s not clear that it is useful to represent the disease course.

Alan Ruttenberg — curator note in the ontology file, 2012

They saw the hole fourteen years ago and were not sure it was worth filling.

Other absences, each confirmed by direct search of the release file. There is no class for:

• dental plaque — named inside the text of the caries definition, with no formal representation of its own
• biofilm, calculus
• fluoride, remineralization
• xerostomia

Everything on the protective side of the equation — the reasons teeth do not decay — is missing. The model runs in one direction only.

Why I am publishing this

OHD is open, actively maintained, and licensed CC BY. It is also, as far as I can tell, invisible to practising dentistry. There is no interactive browser for it anywhere on the public web — the standard ontology portals render it as generic term pages with no sense of the shape.

The shape is the interesting part. What a field has bothered to formalise, and where it stopped, is a reasonably honest map of what that field believes it understands.

Dentistry has written down, rigorously, what a cavity is. It has not written down what a cavity does over time. Those are different problems, and the second is not a matter of adding more terms.

Three routes worth walking

From the top

Start at entity and descend. You pass through pure philosophy — continuant, independent continuant, material entity — then a long stretch of borrowed anatomy before anything recognisably dental appears. That distance is the reuse principle made into terrain.

Sideways from the lesion

Find carious tooth lesion and use the inward relations — the classes that point back at it. That is how you find what a thing participates in, rather than what it is.

Into the dead ends

1,066 of the 1,776 classes have nothing beneath them. Most of an ontology is edge, and the edges are where the questions are.

---

References

1. Duncan WD, Thyvalikakath T, Haendel M, Torniai C, Hernandez P, Song M, Acharya A, Caplan DJ, Schleyer T, Ruttenberg A. Structuring, reuse and analysis of electronic dental data using the Oral Health and Disease Ontology. Journal of Biomedical Semantics. 2020;11:8. doi:10.1186/s13326-020-00222-0
2. Oral Health and Disease Ontology, source repository. github.com/oral-health-and-disease-ontologies/ohd-ontology
3. OHD entry, OBO Foundry registry. obofoundry.org/ontology/ohd
4. Arp R, Smith B, Spear AD. Building Ontologies with Basic Formal Ontology. MIT Press, 2015.

Derived from the Oral Health and Disease Ontology by Alan Ruttenberg, Bill Duncan and Titus Schleyer, used under CC BY 4.0. Original development was funded by the University at Buffalo School of Dental Medicine and NIDCR grants 5R21DE021178 and 5R03DE023358. Class counts, relation counts and the absences stated here were verified directly against the ohd-full.obo release file.