Blood test predicts tumor response to treatment
June 2, 2026
Blood test predicts tumor response to treatment
At a Glance
- Researchers identified distinct cellular neighborhoods common to different tumors, some of which correlate with treatment response.
- A blood test to analyze these neighborhoods could lead to more effective personalized cancer therapies and improve treatment monitoring and outcomes.
Cancer cells live within complex ecosystems of different cell types. The makeup of the tumor environment can affect cancer progression and whether certain therapies will work. Identifying and characterizing distinct cancer ecosystems can be challenging. Existing methods for doing so either ignore spatial information or are unable to integrate data from diverse sources.
An NIH-funded research team, led by Drs. Aadel Chaudhuri at the Mayo Clinic and Aaron Newman at Stanford University, developed an AI tool for profiling cellular ecosystems that have distinct locations and cell compositions, called spatial ecotypes. They used this framework to identify spatial ecotypes common to various cancers. They also looked for associations between these ecotypes and cancer treatment response. Results of the study appeared in Nature on May 6, 2026.
The framework can accommodate different sample types and genomic data from different platforms. The researchers first looked at data obtained using a technique called spatial transcriptomics. This maps local gene activity within a group of cells. They assembled spatial transcriptomic data from more than 100 human tumors of various types. They also compiled data on gene activity from more than 10 million individual cells from the same tumor types.
From these data, the machine learning framework identified spatial patterns of gene activity that were either common across cell types or specific to one cell type. The team identified nine spatial ecotypes common to the cancers included in their analysis based on these patterns. Activity in certain characteristic genes distinguished the ecotypes from each other. Some ecotypes occurred more often near the tumor edge while others occurred more often in the core.
The levels of six ecotypes were linked with cancer survival. In particular, two ecotypes, labeled SE7 and SE8, were associated with a favorable response to a cancer treatment called immune checkpoint inhibition. Another ecotype, SE4, was associated with resistance to this treatment. These ecotypes were better than established biomarkers at predicting treatment response.
The researchers then examined whether a blood test could detect the levels of various ecotypes. Blood tests are less invasive than tumor biopsies, and DNA shed by tumors circulates freely in the blood. The team focused on chemical modifications to DNA called methylation, which affect gene activity patterns. They found that different ecotypes were associated with distinct methylation patterns in circulating DNA. They developed another AI tool that could infer the levels of the different ecotypes from these methylation patterns.
To show the usefulness of such a test, the team analyzed the ecotypes in blood samples from 78 melanoma patients being treated with immune checkpoint inhibition. Elevated levels of ecotypes SE7 and SE8 were associated with future clinical benefits from treatment and longer survival. Higher SE4 levels correlated with treatment resistance and shorter survival.
The findings reveal that spatial ecotypes, which could be measured using blood tests, are fundamental features of the tumor microenvironment. Thus, they could be used to track and predict treatment responses without the need for biopsies. The researchers hope that this knowledge could help health care providers determine which therapies will be most effective for a given patient.
“If a patient isn't going to respond, that's time we could be using a different treatment,” Chaudhuri says. “Better upfront decision-making can directly improve outcomes.”
—by Brian Doctrow, Ph.D.
Related Links
- Boosting the immune response to brain cancer
- AI tool predicts response to cancer therapy
- Identifying immune cells for personalized cancer immunotherapy
- Fiber in diet linked to cancer immunotherapy response
- Artificial intelligence and your health: how computers are helping medicine
- Biomarker testing for cancer treatment
- Immunotherapy to Treat Cancer
- Artificial Intelligence (AI)
References
Non-invasive profiling of the tumour microenvironment with spatial ecotypes. Zhang W, Brown EL, Usmani A, Earland N, Kang M, Olelewe C, Viswanathan A, Chauhan PS, Steen CB, Jeon HS, Avagyan S, Alahi I, Semenkovich NP, Schwab JC, Sachs CM, Qaium F, Harris PK, Cai Q, Gentles AJ, Knight J, Graham RP, Bacchiocchi A, Lucas PC, Fields RC, Sznol M, Halaban R, Chen DY, Chaudhuri AA, Newman AM. Nature. 2026 May 6. doi: 10.1038/s41586-026-10452-4. Online ahead of print. PMID: 42092150.
Funding
NIH’s National Cancer Institute (NCI); Stanford Bio-X; Research Council of Norway; National Science Foundation; Stanford University; Cancer Research Foundation; V Foundation for Cancer Research; Alvin Siteman Cancer Center; Melanoma Research Alliance; Virginia and D.K. Ludwig Institute for Cancer Research; Chan Zuckerberg Biohub.
