A new map of lung cancer biology
Researchers are proposing a “Google maps” style approach to cancer treatment, using artificial intelligence and spatial biology to chart non-small cell lung cancer (NSCLC) at a cell-by-cell level. In a study spanning 234 NSCLC patients across Australia, the United States and Europe, Associate Professor Arutha Kulasinghe from the Frazer Institute at the University of Queensland led a team that mapped the tumor microenvironment in unprecedented detail. The goal is to predict how individual tumours will respond to different therapies, and to guide treatment choices with far greater precision than traditional trial-and-error methods.
How the approach works
By integrating machine learning with high-resolution spatial data from tumor samples, researchers can identify which regions of a tumor are likely to respond to a given therapy and which are resistant. This cell-level mapping creates a molecular geography of cancer, showing where immune cells interact with cancer cells, where signal pathways drive growth, and where therapeutic agents may fail. “Using AI and spatial biology, we mapped NSCLC, cell by cell, to understand and predict its response to drug treatment,” said Dr. Kulasinghe.
Implications for treatment decisions
The team envisions a diagnostic workflow that delivers a treatment roadmap for individual patients. Rather than applying a one-size-fits-all regimen, oncologists could select therapies that target the most vulnerable tumour regions while avoiding those unlikely to benefit the patient. As Professor David Rimm of Yale University noted, this work could lay the groundwork for a new diagnostic test to optimise treatment choices in lung cancer.
Why this matters for lung cancer care
Lung cancer remains the world’s leading cause of cancer death, with approximately 1.8 million deaths annually. NSCLC accounts for about 85% of cases, and immunotherapies—while transformative for some patients—cost between $400,000 and $500,000 per patient per year and are effective in only about 20-30% of cases. They also carry risks, including potentially life-threatening immune-related toxicities. The new mapping approach aims to shift the balance toward patients who are most likely to benefit, reducing exposure to unnecessary toxicity and cost.
From lung cancer to broader oncology
Dr. Kulasinghe emphasised that the framework is not limited to NSCLC. “By integrating data on the molecular geography of cancer and machine learning techniques, we can improve treatment decision-making and outcomes for lung cancer patients. This same approach can also be used to inform treatments for other malignancies where immunotherapies are used, for example melanoma, head and neck, and bladder cancer.”
Looking ahead
While the concept is promising, translating these insights into routine clinical diagnostics will require robust validation, standardisation of sampling, and collaboration across institutions. If successful, the approach could become a cornerstone of precision medicine—helping clinicians choose the right drug for the right patient at the right time, and sparing others from ineffective therapies and their risks.
