Diagnoses & personalised medicine - it’s all in the genes
The need
Correctly diagnosing a condition and its most effective treatment is not always straightforward. Sometimes diagnosis and treatment can be a process of trial and error, causing additional stress and frustration for patients and their families, as well as the clinicians treating them.
The solution
A research team at the ʹڲƱ School of Biomedical Engineering is developing artificial intelligence and machine learning tool sets to analyse genetic variations and other biomarkers for different diseases and disorders. This could provide much faster, more accurate diagnoses and support personalised medical treatments based on a patient’s genetic profile.
Imagine going to a clinic, undergoing a blood test or a facial scan, and having even a rare condition accurately diagnosed within minutes. Then imagine being prescribed a treatment plan customised to you and your circumstances, and offering the greatest possible chance of controlling or curing your condition, in the fastest possible time.
That’s the exciting vision that drives the work of Dr Hamid Alinejad Rokny and a multidisciplinary team of 13 researchers at the ʹڲƱ BioMedical Machine Learning Laboratory.
In pursuit of that powerful – and achievable – future, Hamid and the team are immersed in the development of tools that will use artificial intelligence (AI) and machine learning to accurately identify the biomarkers for a vast range of diseases and disorders.
But diagnosis is just a first step, explains Hamid, who leads the ʹڲƱ BioMedical Machine Learning Lab.
“We can use these as diagnostic tools, but the next stage is about identifying the biomarker and then developing new, personalised treatments based on that biomarker,” he says.
This is about the future. It is about the integration of machine learning and AI for disease prediction
Practical advantages of AI-assisted diagnosis
With a background in systems biology, AI and machine learning, Hamid is keen to see diagnostic genomic tools become common in the Australian health system. To explain their potential for impact, he gives the example of a young child with a notable, observable feature or ‘phenotype’.
“We don’t know if it is an autism spectrum disorder, if it is a chromosomal disorder, if it is hyperactivity – and each of them has a different management plan. It's quite important to find out which type of disease they have,” he explains.
Already, NSW Health Pathology is testing ʹڲƱ-developed AI packages that can pinpoint some genetic conditions in less than 30 minutes. Hamid anticipates that in the next few years, these tools will be commercialised and in routine clinical use across the world.
A current collaboration with Sydney Children’s Hospital and NSW Health Pathology is looking at more than 10,000 individuals to explore a rare genetic disorder that includes a particular facial phenotype.
“We are the power of AI that can look at the face, then associate the facial dimorphism to a specific genetic abnormality without doing any genomic sequencing, without doing any blood test,” says Hamid.
“This is about the future. It is about the integration of machine learning and AI for disease prediction.”
The promise of personalised medicine
In another project the ʹڲƱ team has developed a new AI-informed approach to identifying and quantifying the type and number of bacteria present in the gut. How the gut microbiome contributes to diseases including obesity, colorectal cancer and even depression is a rapidly expanding area of research. By picking up on key biomarkers, the new tool will be able to identify the risk or the presence of specific diseases. It could also be used to inform personalised treatment plans, and individualised plans for precision nutrition.
“Personalised medicine is quite important,” adds Hamid. “You can have one tablet, say, for hundreds of different individuals with breast cancer -- but this small tablet might not work for 10 of them. Our main goal is to use a patient’s genetic profile, and patient-specific clinical information, to identify patient-specific treatment.”
Hamid acknowledges it’s a big conceptual change – and one that requires significant upfront investment.
“When you look at the technology, it’s very expensive to start with. But I think in the future, we can do personalised medicine much, much cheaper than we are doing it at the moment."
An interdisciplinary approach
When the it mapped about 3 billion nucleotides, the ‘bases’ of DNA, which combine into various sequences within our genes.
Making sense of that data requires researchers from many disciplines. The team Hamid works with includes mechanical engineers and others with backgrounds in computer science, pure mathematics, genetics, biology and medicine. There are also experimentalists and collaborators from industry.
“I strongly believe we don’t have a single research area these days,” says Hamid. “I want to have an interdisciplinary team so we can have different ideas from different disciplines. I’m in the middle. A combination of the technology and medicine – that’s what is interesting to me.”