Australia launches machine-learning centre

Dna sequence
A DNA sequence: multiplied by thousands, it may open the door to personalised medicine.
Credit: Science Photo Library — PSIEKT

When it made the cover of Time magazine 16 years ago, the human genome project promised a new era of precision medicine: each person’s health care would be customized to their genetic blueprint.

Some critics have quipped that so far, the impact on medicine has been minor.

Now a new collaboration aims to fulfil the promise. Launched this week by geneticists at Sydney’s Garvan Institute and computer scientists at Victoria’s Deakin University Centre for Pattern Recognition and Data Analytics (PRaDA), it will use machine learning to decode individual human variation.

“We are taking the first steps into a new world,” says John Mattick, the Garvan’s executive director.

“It will fundamentally change the way discoveries are made, the way the health system is organised, and have a major impact on the national economy.”

The Human Genome Project has certainly delivered when it comes to diagnosing rare, single gene diseases. But the common stuff – the cancers, the heart disease, the mental illness – turns out to be complex.

These diseases turn out to involve hundreds of genes acting together. And forecasting just how they act is, so far, beyond current algorithms. Whether or not we get schizophrenia for instance, is about 80% determined by our genes. But so far, scanning the genome is unable to deliver a reliable prediction of who will get the disease.

Recent studies have identified 108 genes, and these predict only 4% of the risk.

Our DNA may be complex but it is decodable. Mothers do it all the time – and with high fidelity. Look at identical twins; given the same DNA, mothers largely produce extremely similar individuals.

For scientists to decode genomes with that same fidelity will probably take the decrypting powers of machine-learning algorithms. It’s an approach pioneered in 2013 by Human Longevity, a San-Diego-based company co-founded by Craig Venter, leader of the privately-funded group that first read the human genome.

Now Australia is getting into the act. Mattick has joined forces with computer scientist Svetha Venkatech, who heads Deakin University’s PRaDA.

Their collaboration – the Garvan-Deakin Program in Advanced Genomic Investigation (PAGI) – was formally sealed on Tuesday .

Both institutes have form. Garvan, having so far read the genomes of 14,000 individuals, and able to process around 1200 per month, is now one of the largest DNA sequencing centres in the world. And as a scientist, Mattick is renowned as a pioneer of attempts to decode the 97% of our genome known as junk DNA.

PRaDA has some great success stories applying machine learning to diverse problems. Their global spin-offs include iCetana, intelligent video surveillance software that detects potential security threats in large data sets.

Another, TOBY, is an app that provides tailored learning for children with autism.

Can Australia compete with the might of a $300 million company founded by Craig Venter?

Absolutely, believes Venkatech. “They [Human Longevity] haven’t solved the problem,” she says.

Neither is she daunted by the competition. She points out the same challenges applied when they entered the security space, yet they came up with “a novel twist” that delivered iCetana.

“We’ll find a twist here. Biology is very complex; it won’t be solved by one group.”

Another strength for Australia is its large clinical data sets, such as Monash’s ASPREE trial that combines 16,000 patients’ DNA with their detailed clinical symptoms. So far, 2000 genomes have been sequenced, and Mattick is keen to extend the collaboration.

“Subject to agreements … this may be the first exemplar project of the next generation of data-driven biomedical discovery and healthcare transformation,” he says.

University of Queensland Geneticist Peter Visscher is excited by the new announcement. “ I believe data science is the way forward; these are powerful complementary approaches,” he says.

And as for genome hype, Mattick paraphrases a quote from Bill Gates: “New discoveries are overhyped in the short-term and underestimated in the long term.” In Mattick’s words, “Precision medicine is ready for the street now.”

Please login to favourite this article.