Enhancing Prostate Cancer Genomics with AI-Powered Morphometric Analysis

By Myriad Oncology^™

As AI becomes more widely used in medicine, prostate cancer is emerging as one of the areas where its impact is most tangible. In pathology, AI can assist in grading prostate cancer more consistently, reducing variability, and revealing subtle morphological features in diagnostic tissue that may be missed by the human eye.¹

Trained on large datasets, AI algorithms can also help predict responses to common therapies, such as radiation or androgen deprivation therapy (ADT), and identify patients who may benefit from treatment intensification or de-escalation at different stages of disease. Increasingly, these tools are being integrated into multiple points of clinical management, making AI not just a buzzword, but a clinical reality.

As AI earns its place in the prostate cancer toolkit, an important distinction is clear: AI and genomics are not interchangeable. Each provides unique insights, approaches clinical questions from a different perspective, and captures information that the other cannot. The most powerful approach is one that combines both.

AI Is Individual Pattern-Driven. Genomics Is Individual-Biology Driven.

At its core, AI is a tool for identifying patterns. Whether analyzing thousands of digitized biopsy images or correlating clinical features across large patient populations, AI works by recognizing trends that even the trained pathologist and clinician might miss.¹ Its strength lies in scale, detecting patterns and relationships that are imperceptible to human eyes due to sheer data volume and/or are too complex for any human to process.

Put simply: AI and genomics each uncover patterns, but from different dimensions of the disease. AI interprets complex sub-visual and clinically relevant signals across large datasets to identify morphometric features and phenotypic patterns mapped to underlying known or yet unknown genomic mutations, RNA expressions and signaling pathways activation that are linked to treatment outcomes. Genomics captures molecular activity within a specific tumor, revealing its biological behavior based on the expression of certain genes with known functionalities. Both approaches offer powerful, individualized insights while addressing various clinical questions. And in both cases, their accuracy depends on the quality and diversity of the data behind them.

This is why AI excels at tasks like recognizing histological patterns, predicting recurrence risk based on diagnostic slide images, or stratifying patients into risk categories for clinical trials. In fact, AI tools can also predict outcomes like biochemical recurrence or metastasis and guide treatment decisions such as whether to add ADT to radiation.² These tools can even help guide treatment acceleration with anti-androgen therapy and chemotherapy in both non metastatic and metastatic patients.

On the other hand, genomic tools like the Prolaris^® Prostate Cancer Prognostic Test offer a complementary perspective by examining functional tumor biology. Through real-time analysis of specific gene expression in a patient’s tissue, these tests provide individualized insight into the tumor’s intrinsic behavior and aggressiveness, capturing molecular signals that may not.

Why Representation Matters in Model Design

As AI models become more widely used in clinical settings, accuracy across groups remains a central concern. Fortunately, there has been encouraging progress. Some AI tools have demonstrated equivalent performance across racial groups by training on diverse datasets.

PATHOMIQ is an example of a company that applies AI across multiple cancer types and patient groups³, while also taking steps to validate its algorithms in demographically diverse cohorts. This kind of inclusive model design is essential, not just for fairness/inclusivity, but for reliability in real-world care.

Genomics, too, plays a critical role in promoting equity. Since genomic expression reflects tumor behavior regardless of external clinical factors, it can help level the playing field when there are disparities in access, treatment timing, or follow-up. Together, AI and genomics offer a path forward, if they’re implemented thoughtfully and transparently.

Rethinking Independence in AI Prostate Cancer Testing

Some tools in the prostate cancer space market themselves as “independent” decision aids, providing a score that sits apart from other clinical factors. While that may sound appealing, it often requires clinicians to reconcile these AI scores with separate tools like CAPRA⁴ and clinicopathology.

By contrast, the Prolaris Test integrates both genomics (cell cycle progression) and clinical features, like PSA and Gleason scores, directly into a clinician-facing risk report that already incorporates CAPRA⁴ and other clinical parameters, giving a more holistic view of patient risk. The test is designed to support, not supplant, clinical expertise by consolidating key insights into a single, actionable report.

While the PATHOMIQ prostate adenocarcinoma (PRAD) score can be an independent predictor of outcome, it also includes a cancer cell morphology-based Gleason scoring system along with tumor microenvironment variables in its algorithm. In addition, it is flexible enough to integrate clinical parameters to offer a highly accurate predictive test.⁵ Combining these tests is a more powerful test than each individually.

BCR: biochemical recurrence

Applying the Right Tool to the Right Question

“I don’t think [we] should look at these biomarkers as ‘which is best’ or a race to a finish line, but [we] should look at them as ‘which is the right tool for the question I am asking?‘” said Jonathan Tward, MD, Ph.D., of the University of Utah.⁷

For instance, an AI prostate cancer tool may flag a patient’s biopsy as potentially undergraded or identify subtle features that suggest a higher likelihood of recurrence.² That’s a cue to dig deeper. A genomic prostate cancer test like the Prolaris Test can then assess the tumor’s biological aggressiveness, helping clarify whether this patient may receive more aggressive therapy or continue with a conservative approach.

This layered approach becomes even more powerful in cases of uncertainty, such as intermediate-risk disease, where treatment decisions are highly individualized. AI can identify risk signals based on clinical patterns and pathological features, while genomics adds a direct readout of tumor behavior through gene expression. Used together, they offer a more complete picture, transforming what might be two separate inputs into a more confident, unified decision.

Bringing It Together: Genomics Enhanced by AI

When paired with genomics, AI becomes a precision amplifier, not a standalone answer. What comes next isn’t AI versus genomics. It’s AI plus genomics, together, each supporting the other and giving providers additional insights like never before.

Both AI and genomic tests like the Prolaris Test are evolving. Soon, the Prolaris Test will integrate AI from PATHOMIQ to provide urologists and radiation oncologists with molecular and AI-driven insights to inform decisions both at biopsy and initial treatment.⁸ It’s not a replacement; it’s an upgrade. And it’s designed to continue supporting physicians with a comprehensive, clinically validated report that respects their judgment and enhances decision-making.

Because ultimately, the role of technology in cancer care isn’t to take the wheel. It’s to offer better maps.

References:

1. Solan M. Artificial intelligence in prostate cancer. Harvard Health Publishing. August 1, 2024. https://www.health.harvard.edu/mens-health/artificial-intelligence-in-prostate-cancer
2. Weight C, host. How AI is Predicting Prostate Cancer Recurrence. Cancer Advances [podcast]. Cleveland Clinic; June 19, 2025. https://my.clevelandclinic.org/podcasts/cancer-advances/how-ai-is-predicting-prostate-cancer-recurrence
3. PathomIQ. Accessed July 30, 2025. https://pathomiq.com/
4. CAPRA: Cancer of the Prostate Risk Assessment; predicts an individual’s likelihood of metastasis, cancer-specific mortality, and overall mortality across multiple treatment approaches.
5. Fay M, et al. 2025. Artificial Intelligence-Based Digital Histologic Classifier for Prostate Cancer Risk Stratification: Independent Blinded Validation in Patients Treated With Radical Prostatectomy. JCO Clinical Cancer Informatics 9:e2400292. doi:10.1200/CCI-24-00292. PMID:40532127; PMCID:PMC12184973.
6. Canter, C et al. Biopsy-delivered cell cycle progression score outperformed pathological upgrading or upstaging in predicting biochemical recurrence after surgery. Poster session presented at: Western AUA; 2018 October 28; Maui, HI
7. Myriad Oncology. Provider perspective: how to choose the right tool for precision prostate cancer care. Myriad Genetics. Published May 15, 2024. https://myriad.com/urology/blog/provider-perspective-how-to-choose-the-right-tool-for-precision-prostate-cancer-care/
8. Myriad Genetics, Inc. Myriad Genetics partners with PATHOMIQ to add artificial intelligence technology platform to its oncology portfolio. Myriad Genetics Investor Relations. Published February 24, 2025. https://investor.myriad.com/news-releases/news-release-detail/26026/