A discussion at the MACSYS retreat ends just weeks later with three MACSYS researchers taking part in a global hackathon – and finishing on a top-three team in the world in their category.

The three researchers, Dr Sophie Wharrie (University of Melbourne) and Dr Ke Ding and Jingsen Yang (ANU), joined an international team to take part in the recently completed Evolved 25 AI x Bio Global Sprint Hackathon, an intensive 10-day event attracting more than 100 teams worldwide.

“We first heard about this hackathon during the retreat and thought it would be a great opportunity. We stayed up late to pull together an application that aligned with our interests and submitted it right then and there,” Ke recalled.

The trio teamed up with Diatom Biolab, a group of startup founders, wet-lab scientists, bioinformaticians and AI engineers from Australia, China and the United States. The team was selected as one of just 10 groups worldwide to compete in one of the three hackathon challenge areas: “Integrative Intelligence in Biomedical Systems Modelling.”

Over the intense 10-day sprint, the group explored synthetic lethality prediction – a promising strategy in cancer therapy that aims to identify pairs of genes whose joint disruption kills tumour cells but leaves healthy cells unharmed. The MACSYS trio led the modelling work to try to predict gene combinations that might work.

“Finding these pairs experimentally is extremely difficult when you consider there are about 200-million possible combinations,” said Sophie.

The team took an AI-Machine learning approach known as virtual cell modelling.  These are computational models for predictive biology trained on millions of biological data points.

“In just 10 days, we were able to get a basic model working that predicted gene pair possibilities, including some that hadn’t been tested before,” she said.

Wet-lab members of the team could then run experiments on those predictions, feed the results back into the algorithms, and iteratively improve the system, making it virtuous cycle of modelling and experimentation.

The 10-day sprint meant that team had to work across time zones. That turned out to be both a challenge and an advantage.

“One memorable aspect was how we worked almost around the clock,” Ke said. “Australia-based members would often work until midnight, then hand everything over to the U.S.-based teammates who were just starting their day.”

Although the sprint focussed on synthetic lethality, the MACSYS team saw the event as an opportunity to advance their broader research goals. “Our main interest wasn’t in the challenge task itself,” Ke said. “We’re focused on how virtual cell models could support MACSYS’ ambition of building a Whole Cell Model. This was a great way to explore questions central to that work.”

Sophie echoed that sentiment. “I’m excited to take what I learned into the rest of my research. The sprint gave me a chance to quickly test and validate some of my intuitions about how these models behave.”

For Jingsen, taking part in a global sprint like this during his PhD was a rare and valuable opportunity.

“You always learn something from others,” he said, “whether it’s getting inspiration from someone in my own discipline or discovering new approaches  from teammates with completely different backgrounds.”

The team’s performance earned them one of only nine invitations to present their results at NVIDIA’s headquarters, where one of the U-S based team members will share the team’s work with investors, potential partners and NVIDIA researchers.  It’s a strong finish to an intense global challenge.

FEATURE IMAGE:  Sophie, Jingsen & Ke at the MACSYS Research Fellow/Student Workshop at The University of Melbourne during the MACSYS Retreat Week