Two Sides of the Same Coin in Prostate Cancer Care

Prostate cancer is one of the most common cancers in men, but every case is unique. Some tumors sit quietly for years, while others spread quickly and need aggressive treatment. Doctors have relied on blood tests, biopsies and scans to figure out which patients need more treatment. These tools help, but they are not perfect. 

That’s why biomarkers are beneficial. Two different types can give extra clues about how a tumor might behave.  

• Genomic classifiers look at the tumor’s RNA expression and search for clues linked to aggressive behavior. 
• AI biomarkers use computer algorithms to analyze digital images of your biopsy tissue, along with other clinical information, spotting patterns doctors might not see. 

On their own, each is helpful. But researchers are now exploring how powerful they can be when used together. “This specific topic has been a clinical conundrum for a while,” said Kosj Yamoah MD, PhD, chair of the Radiation Oncology Department at Moffitt Cancer Center and a presenter at the American Society for Radiation Oncology’s annual meeting. 

Reading Cancer Two Different Ways 

For years, doctors have looked at the genomic pattern of a tumor to predict how aggressive it is. Sometimes a tumor looks low risk under a standard definition but shows dangerous molecular changes. That could push doctors to treat the patient more aggressively.  

Other times, a tumor looks calm at the genomic level, allowing doctors to hold back on extra treatment.  

“That’s how we’ve used genomic data,” Yamoah said. “To complement our ability to treat patients but also manage them after treatment.” 

We are hoping that that data can help us truly understand how we can integrate these novel technologies better, and not only for prostate cancer, but for other disease sites that are really beginning to integrate AI into their mainstream.

Kosj Yamoah, MD, PhD

Artificial intelligence is adding a new angle. Computers can study high-resolution images of biopsy slides and find patterns too subtle for the naked eye. 

“AI technology can actually look at just a picture and tell you kind of the same information: How aggressive is this?” Yamoah explained.  

Two Sides of the Same Coin 

This is where the metaphor “two sides of the same coin” comes in. Both tests are looking at the same disease but from different angles. 

Yamoah compares it to going through airport security. The facial recognition scanner makes a quick judgment based on your appearance. Your passport gives a more detailed description of who you are and where you come from.  Both systems are looking at the same person, but in a different way.  

“If we realize it’s actually the same coin, giving us information from two perspectives, perhaps we can understand how to integrate this better,” Yamoah said.  

When Tools Don’t Agree 

In most cases, the genomic classifier and the AI biomarker align, but sometimes using both tools can create a challenge. In 10% to 15% of patients, the tools conflict, leaving doctors with a dilemma.  

“Clinicians are faced with the issue where the genomic classifier might say that the patient is aggressive, but then AI says that patient is not aggressive and vice versa,” Yamoah said. “When they are concordant, no problem. When they are discordant, then we have issues.” 

Don’t miss Vaseem Khatri, MD (@VaseemKhatri) on behalf of Kosj Yamoah, MD, PhD (@KOSJ12) tomorrow as he presents “Genomic Classifier and Multimodal AI Biomarker in Localized Prostate Cancer: Two Sides of the Same Coin” during the oral session. #ASTRO25 #MoffittASTRO25 @ASTRO_org pic.twitter.com/jy6NVp9MoF

— Moffitt Cancer Center (@MoffittNews) September 30, 2025

The Findings 

Researchers did a deep dive into prostate cancer cases to identify markers linked to more aggressive disease. They found that the genomic classifier and the AI biomarker did not match perfectly. At first, it looked like a conflict, but the tools were still pointing to the same risk: The cancer was more likely to spread.  

The difference was in which biological pathway each tool highlighted. By looking at both the overlap and the differences, doctors can get a more complete picture of each patient’s cancer. 

“We are hoping that that data can help us truly understand how we can integrate these novel technologies better, and not only for prostate cancer, but for other disease sites that are really beginning to integrate AI into their mainstream,” Yamoah explained. 

A More Personalized Plan in the Future 

The next step is to figure out how to use these findings in treatment. For example, patients flagged as high risk by some markers may not always benefit from standard antihormonal  therapy. In those cases, doctors could explore other treatments instead of relying on antihormonal therapy alone. 

The long-term goal is to integrate these tools, enabling doctors to make more personalized decisions, not just in prostate cancer, but across other cancers where AI is being introduced. 

“It’s really about moving toward more personalized medicine,” Yamoah said. “Instead of treating every patient the same, these tools give us a way to tailor care and be more precise.”