Two separate research teams have built tools that bring AI cancer treatment prediction closer to the clinic—one that forecasts whether immunotherapy will work using only routine blood tests, and another that identifies tumor cells in brain tissue in under 15 minutes during live surgery. Together, they represent a shift toward faster, cheaper, and more accessible decision-making in oncology, grounded in data patients already generate during standard care.
Why It Matters
Cancer remains the second leading cause of death in the United States, claiming roughly 600,000 lives per year according to the National Cancer Institute. Immune checkpoint inhibitors—drugs that unleash the immune system against tumors—have extended survival for patients with melanoma, lung cancer, bladder cancer, and other advanced malignancies. But these drugs carry serious side effects, significant cost, and a hard truth: in most patients, the cancer either does not respond or eventually returns.
The two FDA-approved tests currently used to predict immunotherapy response—PD-L1 expression and tumor mutational burden (TMB)—are imperfect. Some patients with low scores respond well; others with high scores gain nothing. TMB testing is expensive and often not covered by insurance. Meanwhile, in brain cancer surgery, surgeons face real-time decisions about how much tissue to remove, often without molecular-level feedback until well after the operation ends. Both gaps leave physicians relying on incomplete information at critical moments.
How These Tools Work
SCORPIO: Predicting Immunotherapy Response from Blood Work
SCORPIO is an AI-based prediction model developed by Diego Chowell, Ph.D., of the Icahn School of Medicine at Mount Sinai, and published January 6, 2025, in Nature Medicine. Unlike existing biomarkers, SCORPIO does not require tumor gene sequencing. It is, as Dr. Chowell described it, “constructed using information pulled only from routine blood tests and patients’ medical records.”
The model was built using data from approximately 2,000 patients treated at Memorial Sloan Kettering Cancer Center and then tested against data from nearly 10,000 patients across two large real-world cohorts and 10 clinical trials. It incorporates simple clinical variables—age, sex, body mass index, and standard blood panel measurements—to predict both tumor response and survival following immune checkpoint inhibitor treatment. A total of 21 cancer types were represented, including melanoma, lung, bladder, liver, and kidney cancers.
SCORPIO builds on a predecessor model called LORIS, developed by Eytan Ruppin, M.D., Ph.D., of NCI’s Center for Cancer Research, in collaboration with Dr. Chowell’s team. LORIS showed promise but still relied on TMB as one of its inputs. SCORPIO removes TMB from the equation entirely, making it cheaper and more broadly deployable. As Dr. Ruppin explained: “In many patients, you can consider a few different types of drugs. So, we’d like to know which one a patient’s cancer is most likely to respond to, to help us carefully weigh the benefits versus the risks and side effects.”
Ultra-Rapid ddPCR: Real-Time Genetic Testing During Brain Surgery
In a separate advance published February 25, 2025, in Med, researchers at NYU Grossman School of Medicine developed Ultra-Rapid ddPCR (droplet digital polymerase chain reaction)—a method that measures levels of specific genetic mutations in brain tissue in just 15 minutes. Standard ddPCR takes several hours, making it useless during a live operation. The NYU team cut DNA extraction time from 30 minutes to under 5 minutes, increased chemical concentrations in the PCR process, and sped up thermal cycling.
The tool tests for two mutations—IDH1 R132H and BRAF V600E—commonly found in brain tumors but absent in healthy brain tissue. It detected as few as five cancer cells per square millimeter across more than 75 brain tissue samples from 22 patients.
Faster, cheaper cancer intelligence—pulled from blood work already in your chart and tissue already on the surgical table—may reshape how treatment decisions are made.
Study co-leader Daniel Orringer, M.D., a neurosurgeon at NYU, emphasized the clinical gap this fills: “But gliomas tend not to have sharp margins, and we need technology to help guide resection.” Unlike imaging tools such as intraoperative MRI, Ultra-Rapid ddPCR operates at the microscopic scale where tumor infiltration actually occurs. Dr. Orringer noted that existing tools miss this level of detail because “tumor infiltration occurs on a microscopic scale, below the level of detection of existing tools.”
The Numbers
- SCORPIO accuracy: 72% to 76% in predicting survival across different patient groups over 2.5 years
- SCORPIO training data: ~2,000 patients (initial build); validated against ~10,000 patients total
- Cancer types covered by SCORPIO: 21, including melanoma, lung, bladder, liver, and kidney
- Ultra-Rapid ddPCR turnaround: 15 minutes (vs. several hours for standard ddPCR)
- DNA extraction time: reduced from 30 minutes to under 5 minutes
- Detection sensitivity: as few as 5 tumor cells per square millimeter
- Tissue samples tested: 75+ samples from 22 brain surgery patients
- Mutations detected: IDH1 R132H and BRAF V600E
“That’s a remarkable performance, just using routine blood tests and basic [patient] data.”
— Diego Chowell, Ph.D., Icahn School of Medicine at Mount Sinai
What Comes Next
Dr. Chowell’s team is now working with hospitals and clinics to test SCORPIO prospectively—in cancer patients who have not yet received immune checkpoint inhibitors—although the score will not be used to determine whether patients should receive treatment during this phase. The team is also building a cloud-based platform to make SCORPIO publicly accessible, following the path of LORIS, which is already available on the NCI website.
For Ultra-Rapid ddPCR, the NYU researchers plan to automate the process to make it faster and more practical in operating rooms. They intend to conduct a clinical trial to determine whether real-time genetic feedback during surgery actually improves patient outcomes. Jing Wu, M.D., Ph.D., a neuro-oncologist in NCI’s Neuro-Oncology Branch, noted that while the technology shows clear potential “in making timely intraoperative decisions,” it still needs further refinement before use outside of clinical studies.
Study co-leader Gilad Evrony, M.D., Ph.D., at NYU, signaled broader ambitions: “We hope our technology and future methods will make molecular-genetic information rapidly available in the operating room for many types of cancers.” Both research teams acknowledged that additional data sources—genomic, proteomic, and clinical—could further sharpen predictive accuracy as these tools mature.
What This Means for You
Neither SCORPIO nor Ultra-Rapid ddPCR is available in routine clinical practice yet, and no reader should make treatment decisions based on these early-stage tools. That said, both studies reinforce a theme that matters well beyond the lab: the body’s own data—blood markers, metabolic signals, genetic profiles—is becoming the foundation for smarter cancer care. The better your baseline health data, the more useful these tools become when they arrive.
This is where daily nutrition and metabolic health earn their relevance. Maintaining a nutrient-dense diet, supporting your micronutrient intake through quality vitamins, and reducing chronic inflammation with greens rich in polyphenols are all within your control right now. These choices do not treat or prevent cancer on their own, but they contribute to the metabolic baseline that AI prediction models increasingly rely on. Supporting recovery and stress resilience also matters—chronic inflammation and poor sleep are associated with impaired immune function, and immune function is precisely what checkpoint inhibitors are designed to amplify.
If you want to go deeper on how supplementation fits into a broader health strategy, our guides on whether vitamin supplements are necessary and how to choose the right supplement for your needs are worth your time. As always, talk to your physician about any individualized health decisions—especially anything related to cancer screening or treatment.
The Bigger Picture
AI cancer treatment prediction is moving from academic curiosity to clinical tool. SCORPIO shows that routine blood work—the kind already sitting in your medical chart—can outperform expensive genomic tests at forecasting immunotherapy outcomes. Ultra-Rapid ddPCR shows that molecular-level tumor analysis no longer requires hours in a lab when minutes in an operating room will do. Neither tool is finished, but the direction is clear: faster decisions, cheaper inputs, and wider access. The patients who benefit most will be those whose health data is rich, current, and well-maintained—a case for proactive health management that extends well beyond any single diagnosis.
Sources
- NCI Cancer Currents: Can AI Help Predict Which Cancer Patients Should Be Treated with Immunotherapy? — National Cancer Institute, February 26, 2025
- NCI Cancer Currents: Rapid Genetic Test Could Help Guide Brain Cancer Surgery — National Cancer Institute, May 8, 2025
- SCORPIO study published in Nature Medicine, January 6, 2025
- Ultra-Rapid ddPCR study published in Med, February 25, 2025
- Background statistics: National Cancer Institute — Cancer Statistics