A new AI-powered gene-mapping method from the University of South Australia reveals that cancer is driven by networks of cooperating genes, not just single mutations. The approach could help doctors find new treatment targets, especially for patients without well-known cancer mutations.
Cancer researchers in Australia have unveiled an artificial intelligence method that maps how groups of genes work together to drive tumors, offering a fresh way to find hidden targets for treatment.
Developed at the University of South Australia and published in the journal Royal Society Open Science, the approach shifts attention from single, heavily mutated genes to the complex gene networks that help cancers grow, spread and resist therapy.
Instead of asking which genes are most frequently mutated across patients, the team asked how genes influence one another over time as a tumor progresses.
The AI system tracks those relationships to expose the biology that makes tumors so hard to stop, according to lead researcher Andres Cifuentes-Bernal.
“The system assesses how genes influence each other over time, providing a clearer picture of the underlying biological approaches that enable tumours to grow, spread and resist treatment,” Cifuentes-Bernal said in a news release.
For years, large genome-wide cancer studies have focused on mutations that show up again and again in many patients. That strategy has been successful in flagging major cancer drivers and has led to targeted drugs for some of them. But it has limits.
“Traditional genome-wide cancer studies typically focus on mutations that appear frequently across patients. While this approach has uncovered many well-known cancer drivers, it overlooks subtle or rare genetic changes. Crucially, it also misses the complex interplay between genes that allow malignant cells to gain momentum,” Cifuentes-Bernal added.
The new framework is designed to capture that interplay.
The method reflects a more realistic view of how tumors behave inside the body, according to co-author Thuc Le, an associate professor at UniSA.
“Cancer is not static,” he said in the news release. “It develops through a cascade of dynamic changes. Many genes act together to disrupt normal cell behaviour, but existing methods can struggle to detect that. Our approach is designed to capture that complexity.”
To test their system, the researchers turned to large breast cancer datasets. Using AI tools, they reconstructed how genes interact in tumor cells and identified clusters of genes that appear to cooperate in pushing cancer into more aggressive stages.
The method was able to pick up many well-known cancer genes that are already cataloged in the Cancer Gene Census, an international reference of genes linked to cancer. That helped validate the accuracy of the approach.
Crucially, the AI also flagged genes that are not obviously mutated but still play powerful roles by influencing other genes. Some of these newly highlighted candidates are involved in cell signaling, immune responses and metastasis, the process by which cancer spreads to other parts of the body.
Rather than spotlighting one gene at a time, the method identifies cooperative networks of genes that act together.
“These networks highlight how genes collaborate to collectively push cancer into more aggressive states,” added Le.
That network view could be especially important for patients whose tumors do not carry the famous mutations that many current drugs target. For those patients, doctors often have fewer options and less guidance on which treatments might work best.
By revealing new genes and pathways that help tumors progress, the UniSA framework could point to fresh therapeutic targets or combinations of drugs that disrupt entire networks instead of single genes.
The work also underscores how artificial intelligence is reshaping biomedical research. AI is particularly well suited to finding patterns in huge, complex datasets, such as the thousands of genes that can be active in a tumor at any given time and how their activity changes as cancer advances.
Understanding those patterns is key to seeing the bigger picture of tumor evolution, according to Cifuentes-Bernal.
“Understanding these dynamics gives us a richer view of how tumours evolve,” he said. “It moves us beyond thinking about single-cell mutations and towards a better understanding of the broader biological systems at play.”
While the current study focused on breast cancer data, the researchers say their framework is adaptable. In principle, it could be applied to other cancers and to diseases where gene regulation shifts over time, including neurodegenerative conditions, autoimmune disorders and chronic inflammatory diseases.
Source: University of South Australia