Stanford and Mayo Clinic Use AI Blood Tests to Map Tumor “Neighborhoods”

Stanford Medicine and Mayo Clinic researchers have developed an artificial intelligence framework that uses blood-based liquid biopsies to map the cellular environments surrounding tumors, a step that could expand precision oncology beyond tumor DNA analysis alone. The research was published in Nature earlier this month.

The system combines two machine learning models called Spatial EcoTyper and Liquid EcoTyper. Together, they analyze methylation signals in cell-free DNA from blood samples to reconstruct tumor microenvironments, including immune and stromal cell activity linked to therapy response.

Current FDA-cleared liquid biopsy workflows largely focus on circulating tumor DNA mutations from cancer cells themselves. The Stanford-Mayo approach instead attempts to profile the broader tumor microenvironment, which researchers say plays a major role in determining whether patients respond to immunotherapy.

The researchers identified nine recurring “spatial ecotypes,” or cellular neighborhoods, across 17 solid tumor types, including melanoma, lung, gastric, bladder, and breast cancers. Some ecotypes were associated with favorable immunotherapy responses, while others correlated with treatment resistance and poorer survival outcomes.