Urine Test Detects Prostate Cancer with Greater Precision than PSA

0
1
Urine Test Detects Prostate Cancer with Greater Precision than PSA


Urine Test Detects Prostate Cancer with Greater Precision than PSA
Credit: Dr_Microbe/Getty Images

A new urine-based diagnostic tool for prostate cancer, combining artificial intelligence and spatial gene expression analysis, could offer a more accurate, noninvasive alternative to current screening methods. The test identifies protein biomarkers in urine that reflect the presence and severity of prostate cancer, achieving diagnostic accuracy surpassing that of the commonly used prostate-specific antigen (PSA) blood test.

“There are many advantages to measuring biomarkers in urine,” said principal investigator Mikael Benson, PhD, senior researcher at the Department of Clinical Science, Intervention and Technology, at Karolinska Institutet. “It’s noninvasive and painless and can potentially be done at home. The sample can then be analyzed using routine methods in clinical labs.”

Details of the new diagnostic method were published in Cancer Research. For their study, the researchers from Karolinska Institutet, Imperial College London, and Xiyuan Hospital at the China Academy of Chinese Medical Sciences analyzed messenger RNA (mRNA) activity across thousands of individual cells from prostate tumor samples. Using spatial transcriptomics (ST), the investigators developed computational timelines of malignant transformation called pseudotime (PT) models and then applied machine learning to identify 45 genes associated with cancer grade, chromosomal changes, and pathways relevant to prostate cancer.

In practical application, the diagnostic models using these biomarkers achieved an area under the curve (AUC) of 0.92 in urine samples, significantly higher than PSA-based diagnostics, which typically range from 0.56 to 0.6. In comparison, blood-based models using the same proteins achieved an AUC of 0.69. These findings suggest that proteins found in urine provide more accurate diagnostic signals than those found in blood.

AUC is a measure of the true positive rate (sensitivity) of a diagnostic versus the false positive rate (specificity), rated on a scale of 0 to 1, with 1 being a perfect score (100% sensitivity and specificity).

Currently, screening for prostate cancer usually relies on PSA level tests, which lack specificity and sensitivity. This often results in false positives and unnecessary biopsies or, conversely, missed diagnoses. Additional tests, such as the Prostate Health Index (PHI), have improved performance slightly, reaching AUCs up to 0.77, but still fall short in diagnostic reliability across patient populations.

To validate the candidate biomarkers, the researchers analyzed samples from more than 2,000 patients, including urine, blood, and prostate tissue. They found consistent patterns of overexpression in biomarkers such as SPON2, AMACR, and TMEFF2 across independent patient cohorts. SPON2, for example, showed higher expression in high-grade cancer tissue compared to healthy or benign samples.

The new tool could also help enhance the delivery of more personalized treatment in the future because some of the biomarkers are linked to known drug targets. “This raises the possibility that ST and PT can be used to identify drug targets and that the protein levels of these targets may be used as a companion diagnostic tool to personalize treatment with such drugs,” the researchers wrote.

The research team has made all data and analytical tools publicly available and is preparing for large-scale clinical trials. One such effort is being discussed with Rakesh Heer of Imperial College London, co-author of the study and head of TRANSFORM, the U.K.’s national prostate cancer study, which provides a platform for rapidly evaluating biomarker candidates.

Future steps include validating the diagnostic model in larger, more diverse populations, refining the test for clinical implementation, and exploring its utility in other cancers. The researchers suggest that their framework, which combines spatial transcriptomics, pseudotime modeling, and machine learning, may be applicable across oncology.



Source link