Prostate cancer ranks among the most prevalent and life-threatening cancers in men, yet early detection remains a critical challenge. The disease often presents subtly in its initial stages, and existing diagnostic tools struggle to identify it reliably.
For years, scientists have recognized that cancer originates from complex gene activity within the body, but translating these changes into actionable early diagnostic tools has proven elusive—until now.
A recent international study marks a potential turning point. Researchers from Sweden’s Karolinska Institute have developed a method to diagnose prostate cancer early using a simple urine sample, combining advanced gene analysis, spatial transcriptomics, and artificial intelligence.
This approach may outperform the widely used PSA test, which is often criticized for its lack of precision.
Mapping the Genes Behind Cancer
Every tumor contains thousands of interacting genes, and their activity evolves as cancer becomes more aggressive. However, these genetic patterns vary between individuals and even within different parts of the same tumor, making it extremely difficult to identify a single biological “marker” that consistently signals early-stage cancer.
To overcome this hurdle, the scientists focused on three key insights: tumors exhibit varying degrees of progression from benign to aggressive; spatial transcriptomics can measure gene activity based on a cell’s location in tissue; and the timeline of these changes—known as pseudotime—can be modeled using such data. By integrating these tools, researchers can track how cells gradually transition from healthy to cancerous.
In the study, scientists used data from three prostate cancer studies to build pseudotime models, analyzing gene behavior in thousands of prostate cells across different cancer grades. This allowed them to pinpoint genes most closely linked to cancer progression.
Building Digital Models with AI
Researchers created digital models using the full mRNA activity of every human gene in these cells, mapping gene activity alongside each cell’s location and cancer stage to form a detailed picture of how prostate cancer grows and evolves over time.
Artificial intelligence was then employed to analyze these models and identify useful biomarkers—proteins that indicate the presence of prostate cancer. The team detected these biomarkers not only in tumor tissue but also in blood and urine samples from nearly 2,000 men, ensuring robust and wide-ranging data. Notably, many markers appeared clearly in urine, enabling easy and non-invasive testing.
Machine learning models validated the urine biomarkers’ accuracy, achieving an AUC score of 0.92—a metric measuring a test’s ability to distinguish healthy from diseased patients. This score reflects very high precision, outperforming the PSA blood test currently used in clinical settings.
Why Urine Testing Matters
Dr. Mikael Benson, who led the study at Karolinska Institutet, highlights the advantages of urine-based biomarker testing: “It’s non-invasive and painless and can potentially be done at home. The sample can then be analyzed using routine methods in clinical labs.”
This approach could make early testing more accessible globally. Instead of requiring a biopsy or hospital visit, men could submit urine samples from home, receiving results sooner and without discomfort. The research also suggests urine markers may reveal cancer severity, helping doctors tailor treatments promptly.
A Path Toward Clinical Use
While the results are promising, further testing is needed. Large clinical trials are planned to validate the method, including one through TRANSFORM, a UK national prostate cancer study led by co-author Professor Rakesh Heer of Imperial College London.
“New, more precise biomarkers than PSA can lead to earlier diagnosis and better prognoses for men with prostate cancer,” Benson explains. “Moreover, it can reduce the number of unnecessary prostate biopsies in healthy men.” Biopsies are invasive and carry risks, so fewer false positives would significantly benefit patients.
Conclusion
This approach, which combines spatial transcriptomics, pseudotime models, and artificial intelligence, could reshape cancer diagnosis. Beyond prostate cancer, it may also aid in detecting other complex cancers. Published in the journal Cancer Research, the study’s strength lies in linking gene changes to cancer grade across over 2,000 samples, with urine biomarkers making early detection more practical.
While further real-world trials are needed to confirm the findings, if validated, these urine-based tests could soon be available to doctors and patients worldwide, offering a new era in early cancer detection.
READ MORE
