At the forefront of the war against cancer, researchers at the U.S. Department of Energy’s Argonne National Laboratory are harnessing one of the world’s most powerful tools: the Aurora supercomputer.
With its unprecedented processing capabilities, Aurora is being used to supercharge artificial intelligence (AI) and high-performance computing (HPC) to discover effective cancer treatments –especially for forms of the disease that resist existing therapies.
Aurora is not just another supercomputer, it’s an exascale system capable of performing more than a quintillion calculations per second.
This power allows researchers to process massive datasets, simulate biological systems in extraordinary detail, and train advanced AI models that would be impossible to handle with previous systems.
Thomas Brettin, computational scientist and strategic programme manager at Argonne, explained: “Argonne has developed AI and supercomputing capabilities that are among the best in the world.
“Part of our mission is to apply these capabilities to grand challenges facing the nation and humanity, and cancer is one of them.”
From AI models to real-world cancer solutions
Argonne’s journey into AI-powered cancer research began nearly a decade ago, spurred by a collaboration between the Department of Energy (DOE) and the National Cancer Institute (NCI).
Early efforts led to the development of the CANcer Distributed Learning Environment (CANDLE), a scalable deep learning framework designed specifically for the nation’s growing portfolio of cancer data.
CANDLE paved the way for predictive models that could estimate how tumours respond to different drugs. These models laid a critical foundation for Argonne’s broader strategy: to combine HPC with AI and lab research to fast-track drug discovery.
Their breakthrough came when researchers used the Aurora supercomputer to screen 50 billion small molecules in just 20 minutes – a task that would have taken weeks on older systems. The speed and scale of Aurora allow for a dramatic compression of the cancer drug discovery timeline.
Pushing boundaries with predictive oncology
Building on this momentum, Argonne launched the IMPROVE project in 2021. This initiative set out to evaluate and benchmark the growing ecosystem of AI models designed for cancer treatment prediction.
Partnering with the Frederick National Laboratory for Cancer Research, the project developed a standardised framework to assess which models best predict drug responses across various tumour types.
This work evolved again when Argonne researchers began exploring large language models (LLMs) – AI tools trained to generate new molecules by understanding patterns in biochemical data.
These next-generation models represent a leap forward, shifting the focus from merely evaluating responses to designing new drug candidates.
Targeting the “undruggable”
Now, the team has set its sights on one of the most difficult challenges in cancer research: targeting undruggable proteins – molecules that have long evaded treatment due to their elusive and unstable structures.
With support from the Advanced Research Projects Agency for Health (ARPA-H), Argonne is partnering with the University of Chicago Medicine Comprehensive Cancer Center on a bold initiative called IDEAL.
The goal is to find small molecules that can inhibit cancer-driving proteins that were previously deemed impossible to treat.
This highly interdisciplinary project brings together structural biology, computational modelling, and experimental validation.
Researchers start by identifying target proteins and, if necessary, determining their 3D structure using the Advanced Photon Source (APS), a DOE facility recently enhanced with brighter X-ray capabilities.
Once the structure is mapped, the Aurora supercomputer simulates the protein’s atomic behaviour. These simulations help identify potential binding pockets – tiny structural crevices where therapeutic molecules might attach and disrupt the protein’s function.
From simulation to laboratory testing
After computational screening, promising molecules are handed off to laboratory teams for testing. Scientists then validate the AI predictions by observing how the compounds interact with the protein or affect tumour growth in experimental models.
This loop of simulation and real-world testing dramatically accelerates the traditional drug discovery pipeline.
What makes this approach groundbreaking is its ability to tackle long-standing roadblocks in oncology. Many of the proteins targeted by IDEAL are intrinsically disordered, meaning they lack a fixed structure. This is one reason why they’ve been labelled undruggable for years.
But with the Aurora supercomputer and AI tools, researchers are uncovering ways to manipulate even these elusive targets.
Why the Aurora supercomputer matters
The Aurora supercomputer isn’t just a powerful machine. It’s a catalyst for scientific transformation. By drastically reducing the time needed to screen potential drugs and simulate protein behaviour, Aurora helps researchers zero in on viable treatments faster than ever before.
This convergence of AI, HPC, and experimental science could usher in a new era in oncology, offering hope for patients whose cancers have resisted traditional therapies.
