A new and powerful Australian artificial intelligence model can comb through data from confidential pharmaceutical trials – both failed and successful – to help big pharma companies avoid wasting their resources on already investigated dead ends. Australian start-up Qubigen developed the AI algorithm that can learn from confidential medical, chemical and biological data and help shorten drug discovery times without compromising intellectual property ownership. The model has already provided promising leads to new pharmaceuticals.
Winning gold in the Medical category of the Financial Review BOSS Most Innovative Companies awards, Qubigen’s AI tool can learn from a trove of pharmaceutical lab investigations, all without copying or sharing the data.
Qubigen founder Jonathan Hall says the AI model is a shortcut through thickets of time-wasting work. It can take many years and billions of dollars to develop a useful vaccine or pharmaceutical, but the Qubigen model can eliminate avenues of inquiry that have proved to be dead ends, and mine the data from game-changing discoveries as a step to further developments.
“These drug companies spend years and years and billions of dollars testing all kinds of molecules,” he says. “Most of those attempts fail. They just churn through so many attempts, attempt after attempt, and all that useful knowledge of the failed attempts can never leave the lab. So the next experiment has to start almost from scratch.”
Qubigen’s patented Federated AI platform for drug discovery and design (FedAIDD) builds AI engines that can learn from confidential pharmaceutical data without moving or exposing it. With this platform, according to the company, pharmaceutical organisations can build tailored, high-performance models to refine and improve their drug discovery work.
“Federated”, Hall says, simply means joined together – so the Qubigen AI model can link sensitive medical chemistry or pharmaceutical data that has never been shared publicly.
The Qubigen AI model absorbs information and learns from these confidential lab results, but it doesn’t reveal the foundational data to anyone, Hall says. Each company that allows Qubigen’s AI model to search through its lab results is provided with a take-home AI model, “tweaked to their own specific data, so they get their own specific molecules”. It is the first commercially deployed system, he adds, to give clients complete ownership and control of their AI models behind their own firewall.
“We’ve got this powerful method … a new kind of AI that can use all the new types of AI knowledge that people are developing now.”
With a background in computational physics, Hall was an academic at the University of Adelaide (now merged with the University of South Australia and renamed Adelaide University) before he left the university in 2016 to launch a medical imaging analysis start-up, Life Whisperer AI.
This start-up used AI to improve the chances of successful in-vitro fertilisation, assessing the health of human eggs and embryos by analysing images securely kept in hospital and clinic files. Now sold to an international company, Life Whisperer AI could assess confidential ovum and embryo images to determine viability without the hospital or clinic actually sharing the data.
“We’ve got this powerful method that is really underpinning the whole thing: a new kind of AI that can use all the new types of AI knowledge that people are developing now,” Hall says. “This is a new way of being able to use data without actually moving or seeing it, and keeping data sovereignty in a day and age where people care a lot about where the data sits.”
Qubigen was launched in the pharmaceutical sector in 2024 after Hall and his then company merged with the Melbourne Group, which provided some seed capital. To date the company has managed as many as 20 projects with a range of world-leading institutes and industry groups.
One promising line of inquiry has been in the field of neurology, which as a whole has had fewer breakthroughs than areas such as oncology. The blood-brain barrier has been an obstacle, limiting which drugs can pass from the bloodstream into the brain and making it more difficult to use pharmaceuticals to treat brain conditions and illnesses.
The Qubigen AI model has made substantial headway here, providing much-needed accuracy concerning blood-brain barrier penetration accuracy, Hall says.
These sorts of solid results are inspiring trust across the sector, Hall says. “Getting good results for people – that’s what really helps.”
Qubigen’s AI model can turbocharge scientific investigation, particularly at the earlier pre-clinical stages of drug discovery, before manual time and effort is required. Toxicity studies, for instance, require a clinical workflow. But Qubigen AI can also help understand toxicity because it has more data to draw on, so scientists are not forced to return to square one when a line of investigation fails.
The company can also license the AI platform so subscribers can use it either for themselves or for their clients. Contract labs with a number of pharmaceutical clients could run a simplified version of the AI model to run their own investigations.
Many technology companies are now riding out the wave of public anxiety about AI, at least partly driven by AI company leaders warning of rogue AI leaping boundaries and making its own decisions.
“It seems like whenever you use the term AI, you have to be careful,” Hall says. “This new AI boom is all around language models and the chatbots.” Qubigen AI is not a chatbot, he adds, but it does use the most modern algorithms.
Convincing people of the intrinsic value of the Qubigen model has required a substantial effort, Hall says, but potential clients are now seeing the value.
Making health start-ups sustainable
Taking silver in the Medical category, the non-profit incubator ANDHealth+ provides support and guidance to digital health start-ups.
The flagship program of Australian connected health commercialisation organisation ANDHealth, ANDHealth+ is a non-dilutive accelerator for scaling up promising digital health companies.
ANDHealth CEO and founder Bronwyn Le Grice, who came from the world of venture capital, says before ANDHealth was launched in 2017, the potential of digital and connected health technologies was poorly understood and underrated in the private sector.
”A different approach was needed, not only with investment, but how we develop the companies, how we commercialise them and turn them into sustainable businesses,” she says.
“The intent was to create a sovereign capability for Australia that demonstrated we could develop and deploy globally relevant digital health technologies that impact global patient populations.”
ANDHealth+ starts to work with digital health companies after they have reached a certain level. The incubator helps with regulatory clearance along with determining levels of customer interest and where the best markets are and why.
Among these companies so far are Oli, which developed a wearable sensor that can predict the likelihood of postpartum haemorrhage; and Osara Health, a digital engagement platform for remote cancer patient care that provides support and guidance, and helps cancer patients return to work if they want. Founded in Sydney, Osara now has branches in the US, Canada and New Zealand.
ANDHealth+ has a focus on companies’ medical claims for technologies, so they are regulated similarly to medical devices, Le Grice says.
“To claim that your AI can rapidly or better diagnose early-stage cancer, you would need to have clinical studies that support that case, just as you would if you were developing a new imaging device,” she says. “We work at that end, where existing regulatory structures and the new technologies collide.”
ANDHealth+ competes for, and wins, significant funding from Australia’s Medical Research Future Fund, the long-term government endowment fund that supports health and medical research.
The incubator identifies the Australian companies with the most potential and funds them in a way that is analogous to hands-on venture capital funding: the funding is in tranches, allocated after certain agreed goals are reached.
ANDHealth+ does not take equity in the companies, Le Grice says.
“But we make the funding contingent upon a very closely managed and agreed project plan, which goes for 18 months over two stages. Our objective is to make sure that when we incubate a company, we really move them from where they were to become more investable and deployable, and of more interest to enterprise customers.”
Scanner detects brain injury at the scene
The bronze winner is EMVision’s First Responder, a backpack-sized portable brain scanner that allows paramedics to quickly diagnose stroke or traumatic brain injury at the scene, and provide the appropriate treatment.
“When a stroke occurs, around two million brain cells die every minute,” says EMVision CEO and co-founder Scott Kirkland, adding that strokes can either be caused by a clot or a bleed in the brain, and the two types of stroke require diametrically opposed treatment.
“There are time-sensitive proven treatments, but they require that differentiation: is it a clot or is a bleed? The earlier these treatments are applied, the more effective they are and the less disability for the patient.”