New research from Memorial Sloan Kettering Cancer Center (MSK) investigates harnessing the power of ferroptosis to spread cell death; reports how an MSK artificial intelligence (AI) model could help improve patient safety; and uses AI to reveal country-specific drivers of global cancer outcomes.
Ferroptosis is a type of cell death that occurs in degenerative conditions and has potential applications in cancer treatment. It is driven by iron-powered lipid damage, which harms cell membranes.
An MSK research team — led by postdoctoral researcher Jyotirekha Das, PhD , and graduate student Saloni Hombalkar , members of the lab of senior study author Michael Overholtzer, PhD — investigated why ferroptosis sometimes kills individual cells and other times spreads as a wave through groups of cells.
They discovered that the spreading type of ferroptosis requires damage to lysosomes — an organelle that recycles cellular waste — and that severe damage causes them to rupture. Enzymes released from lysosomes then help drive necrotic cell rupture. Released iron may additionally amplify damage and potentially affect nearby cells. The scientists also showed that lowering antioxidant defenses, especially by depleting glutathione, shifts cells toward necrosis and enables collective cell death. In contrast, glutathione peroxidase 4 (GPX4) inhibition alone can trigger mixed death types, including apoptosis, which doesn’t propagate well.
The findings could help explain tissue damage in conditions like stroke, as well as suggest ways to steer cancer therapy toward propagating, necrotic ferroptosis that may help eliminate resistant tumor cells.
Read more in Developmental Cell .
Artificial intelligence (AI) is already helping doctors improve radiation treatment and more accurately measure tumors. New efforts could help in another critical area: patient safety.
While infrequent, when medical errors occur, it’s important for hospital teams to understand what went wrong and why — and use those lessons to better protect patients in the future. A new AI-based incident analysis approach developed at MSK was designed to improve the efficiency and consistency of reviewing reported incidents and near-misses.
A team — led by medical physics resident Abbas Jinia, PhD , and overseen by senior study authors Jean Moran, PhD , Director of Division of Radiotherapy Physics, and Anyi Li, PhD , Chief of Computer Service in the Department of Medical Physics — built a system to automate initial reviews of incidents while keeping outputs structured and transparent. The first-of-its-kind approach employs a Human Factors Analysis Classification System used in aviation safety and applies it to healthcare using a large-language model (LLM).
“A central principle of the project is that this analysis cannot be a black box,” Dr. Li says. “For patient safety, an explanation of the classification is not optional. The system is built as an interactive experience, so reviewers can ask why a given contributing factor was assigned and understand the model’s reasoning.”
The model was trained using more than 1,500 mock incidents written by a multidisciplinary team and then tested on 350 real incident reports. The model proved to be 29 times faster than human review alone — sifting the text of reports in mere seconds — and the results agreed with human experts 88% of the time. Disagreements, the authors note, were usually the result of missing clinical context — an area earmarked for future improvements.
“Ultimately, we believe AI-assisted review and classification of incidents can accelerate our learning to improve patient safety. This will allow us to shift our team’s focus to designing safer systems in support of our patients,” Dr. Moran adds.
Read more in npj Digital Medicine .
Despite gains made in wealthier countries, there remain challenging global disparities in cancer care and outcomes.
An international team of researchers — led by Edward Christopher Dee, MD , a radiation oncology resident at MSK and Milit Patel , an undergraduate student at the University of Texas at Austin — built an artificial intelligence (AI) model to better understand how outcomes vary between countries and which factors correspond to improved survival.
Their analysis used widely available measurements including economic strength, availability of universal health coverage, radiotherapy access, workforce composition, health spending, out-of-pocket costs, pathology services, and gender inequality.
“We found three strong, influential factors worldwide: gross domestic product per capita, radiotherapy availability, and universal health coverage — among other complex and interrelated health systems factors,” Dr. Dee says. “And it’s not just how much is being spent, but how efficient that spending is. High out-of-pocket costs also tended to worsen outcomes.”
Read more in the Annals of Oncology .