You might be interested
A complex web of information exists in science and medicine.
To untangle the web, scientists must explain this information to computers. How do they do this and ensure the information represents everyone?
Ontologies may hold the answer.
To be, or not to be?
In philosophy, ontology is the study of reality and existence.
When philosophers such as Descartes and Plato wrote on gods, the universe and life, they were proposing their ontologies.
In science, an ontology is a library of descriptions. Ontologies help classify concepts and their relationships to each other.
Caption: Ontology and neural network papers each year
Credit: Digital Science
Let’s use the human heart as an example.
Matters of the heart
The heart in the FMA ontology is “an organ with cavitated organ parts” and related to veins and arteries.
The Gene Ontology links specific genes to heart growth and function. The Human Disease Ontology describes diseases like heart sarcoma.
A researcher can stack and link these ontologies. These links can predict which genes increase the risk of heart sarcoma and which other organs heart sarcoma affects.
Ontologies are more than technical tools – they’re maps of meaning.
The way researchers draw those maps determines whose knowledge is included and whose isn’t.
Building meaning
Professor Gareth Baynam is a clinical geneticist, policy advisor and the Medical Director of the Rare Care Centre.
Gareth is also a co-founder of the universal Indigenous medical translator Lyfe Languages.
He’s part of the International Rare Disease Research Consortium (IRDRC), which aims to improve the diagnosis of rare diseases in Indigenous people.
Alongside colleagues, Gareth is working to build Indigenous ontologies for health.
This includes Aboriginal ontologies here in Australia. These ontologies may help to close the First Nations health gap by combining ancient knowledge and new technology.
Caption: An artistic representation of the Rare Care Centre Model of Care Ontology
Credit: Ronda Clarke
Combining ancient and cutting-edge
The IRDRC identified the need to link global Indigenous ontologies to the Mondo Disease Ontology.
Mondo unifies genetic, physical traits and disease ontologies. It assesses different traits in both humans and animals to improve disease diagnosis and care.
Most of Mondo’s human ontologies are based on European genetics and Western ideas. This means critical information from other cultures and ancestries may be missing.
“You can think of Mondo as like an atlas,” says Gareth. It’s a set of maps and how those maps connect, as if you drilled holes through the book of maps and connected those maps in different ways.”
Ontologies are especially useful in machine learning. Because computers are very literal, they can struggle to interpret human descriptions of objects and their relationships.
But computers understand ontologies. They can find relationships between genes, diseases and objects. There can be millions of links, too complex for a human to work out alone.
Artificial ontology
Researchers are using ontologies to train artificial intelligence models.
These models are being used to predict breast cancer, future diseases and patient triage.
“As a doctor, when I take a family history, I ask how different people in your family relate – your parents, grandparents, aunts, uncles, cousins,” says Gareth.
“As a Western concept, that’s familiar to many people. But in other cultures, what people describe as an auntie, uncle or grandparent may be different.
“An auntie or uncle may just describe people older than you. We need a way to represent those sorts of details in maps.”
Overlooking or misinterpreting these relationships can have real, measurable impacts on healthcare.
Caption: An example of the gene ontology. Here are the ontology terms to define a biochemical pathway to create sugars.
Credit: National Human Genome Research Institute
A healthier WA
The WA Department of Health assessed our health system in the 2019 Sustainable Health Review. It mapped the challenges and goals of WA health to 2029.
One of the findings was that clinical care only counted 16% towards patient health outcomes, while socioeconomic factors and health behaviours contributed 40%.
“Look at the National Aboriginal Community Controlled Health Organisation ways of improving health and wellbeing. There’s spiritual health, connection to Country, language, family, social and emotional wellbeing,” says Gareth.
“We want a health ontology that’s inclusive and holistic. If we’re missing these elements in our maps, then there are whole segments of health we’re never even going to address.
“The Indigenous ontologies already exist. They have been transmitted through the spoken word for tens of thousands of years.
“We’re trying to connect them to new tools in a way that includes trust and layers of permission. You can’t just extract that knowledge out of someone and use it without permission.”
If Indigenous ontologies don’t become digital, they will remain invisible to new technology and the literal-minded computers.
But there are ethical issues around making this knowledge publicly available.
“There are some key principles in Indigenous data sovereignty – custodianship, access, respect and permission of use,” says Gareth.
“We work with the Elders to make sure we’re using the right words in the right settings. [We’re building] something that is sufficiently trustworthy from an Indigenous health perspective.”
