Photo Credit: iStock.com/Yuuji

Researchers identified urinary biomarkers for PCa that may enable early-stage diagnosis via a simple, noninvasive urine test, according to a recent study.

Employing advanced digital modeling and artificial intelligence, researchers have identified novel prostate cancer (PCa) biomarkers that outperform the conventional prostate‑specific antigen (PSA) and may allow for diagnosis at an early stage through a simple urine sample, according to a study published online in Cancer Research.

“In this study, we aimed to address a general problem in cancer diagnosis, namely that there is a lack of 479 accurate biomarkers that can be measured using routine clinical methods, focusing on PCa,” wrote corresponding author Martin Smelik, PhD student at Karolinska Institutet, and colleagues.

Digital Modeling & Biomarker Discovery

The team applied spatial transcriptomics (ST) and pseudotime trajectory (PT) analysis to identify genes most closely associated with malignant transformation across heterogeneous prostate tumors. Building on prior work that used PT to map bulk transcriptomic changes and nominate EZH2 as a therapeutic target, the researchers shifted focus to biomarker discovery.

ST profiling of 12 tumor specimens produced high‑resolution mRNA maps, from which PT values—reflecting progression toward malignancy—were derived. Genes exhibiting the strongest PT correlation were enriched in hallmark cancer pathways (eg, transepithelial migration, androgen response) and included PCa‑relevant candidates such as SPON2, AMACR, and TMEFF2.

Preclinical Validation

According to the study, high‑PT genes were overrepresented among known PCa drug targets and demonstrated consistent differential expression between PCa patients and controls in independent blood, tissue, and urine cohorts. Immunohistochemical analysis confirmed that SPON2 protein levels increased with tumor grade, underscoring its potential as a tissue‑based biomarker.

Diagnostic Performance in Biofluids

To evaluate clinical utility, the researchers quantified candidate proteins in plasma samples from the UK Biobank (n≈2,000). A multivariate model incorporating these biomarkers achieved an area under the curve (AUC) of 0.69, modestly superior to randomly selected proteins but only marginally better than PSA. Recognizing that local fluids may yield richer signals, the team modeled urinary protein data, attaining an AUC of 0.92 compared to 0.63 for PSA alone.

“Interestingly, in urine, the randomly chosen candidate biomarkers reached an AUC 531 of 0.88,” the authors wrote. “This supports that proteins in local fluids, like urine in the case of PCa, are more informative for prediction models than those from blood and thus, have better diagnostic potential.”

Limitations & Future Directions

Study limitations cited by the authors included variable ST sensitivity for low‑abundance transcripts, incomplete modeling of tumor branching heterogeneity, and a relatively small urine cohort (< 200 samples). Nonetheless, the team noted that biomarker expression correlated robustly with PCa grade and encompassed actionable drug targets, suggesting roles in diagnosis and therapeutic stratification. “Since the analysis of proteins in urine is a tractable diagnostic tool, we propose prospective studies to test the diagnostic potential of these candidate biomarkers in PCa,” the authors concluded.