A predictive artificial intelligence system developed at Memorial Sloan Kettering Cancer Center and Weill Cornell Medicine can determine whether a cancer patient will benefit from immunotherapy by analyzing a single, widely available tumor biopsy slide. The tool, named SCORPIO, works even when the cancer type or the specific immunotherapy drug has never been encountered in its training data.
Researchers describe it as a kind of universal translator that learns the visual language of tumor–immune system interactions and applies that knowledge across cancers and treatments it was not explicitly taught.
How SCORPIO Works
SCORPIO relies on a standard H&E-stained slide of a tumor sample, a routine part of pathology that most hospitals already produce. The AI examines the spatial arrangement of cells on the slide — not merely which cells are present, but how they are organized relative to one another.
The system detects patterns invisible to the human eye. For instance, a tumor with immune cells clustered at its edge may respond differently from one where immune cells have infiltrated the core. This spatial analysis is the key to SCORPIO’s generalizability. Earlier predictive AI models have been brittle: trained on one cancer type and one drug, they fail when either variable changes.
SCORPIO was designed to be broadly applicable. The research team trained the model on a diverse set of tumors and immunotherapy regimens, then tested it on entirely new combinations of cancer types and drugs.
The tool consistently identified which patients would respond to treatment and, crucially, which patients would experience better overall survival. The findings were published in the journal Nature Cancer, based on tests involving more than 5,000 patients across 24 cancer types and 10 different immunotherapy regimens.
Implications for Patient Care
Immunotherapy has transformed cancer treatment, but it is effective only for a subset of patients. Doctors currently have limited methods to predict who will benefit and who will suffer side effects without gain. Dr. Travis Hollmann, a pathologist at Memorial Sloan Kettering and co-senior author of the study, described the potential impact: “This could help identify patients who are most likely to benefit from immunotherapy, potentially sparing others from unnecessary side effects and allowing them to pursue alternative treatments sooner.” The practical significance extends beyond individual patient decisions.
A generalizable predictive tool like SCORPIO does not require separate models for each cancer type or each drug. That means hospitals without the resources to build customized AI could deploy the same tool.
It also addresses a persistent gap in medical AI: many models perform well in controlled studies but fail when faced with data from different hospitals or patient populations. A model that remains reliable across those shifts could bring evidence-based predictions consistently from the lab to the clinic.
The Path to Clinical Use
SCORPIO’s code has been made publicly available so that other institutions can test and validate the system. The researchers caution that the tool needs prospective clinical trials before it can be used in routine patient care.
The next key step is whether SCORPIO holds up in those real-world trials and whether it can eventually guide treatment decisions in the clinic, giving oncologists a fast, affordable test that works across the board. Artificial intelligence has been used in medicine for decades, evolving from rule-based systems dependent on human expert instructions to deep learning that can find patterns in vast visual data. Yet most medical AI tools remain narrowly focused.
SCORPIO represents a step toward generalizable AI — machine learning that can apply what it learns from one set of conditions to entirely new ones without being retrained. Such a tool does not need to see every possible scenario in advance; it extracts underlying principles that hold across different contexts.
For patients, that could mean fewer unnecessary treatments and more informed decisions. (Consult your doctor for medical advice.)
