Note: Single-source report; awaiting corroboration.
NIH-funded researchers have developed a machine learning tool to profile tumor cellular ecosystems—called spatial ecotypes—by analyzing local gene activity in cancer cells. Using spatial transcriptomics data from over 100 human tumors and gene activity from more than 10 million individual cells, the tool identified nine spatial ecotypes common across tumor types. These ecotypes, characterized by specific gene activity, varied in location within tumors, often occurring near the tumor edge or core.
The study found that six of these ecotypes correlated with cancer survival. Two, labeled SE7 and SE8, were linked to favorable responses to immune checkpoint inhibition therapy, while SE4 was associated with resistance. These spatial ecotypes offered better prediction for treatment response than existing biomarkers.
The researchers also explored the potential for a blood test to detect the levels of these spatial ecotypes, which would be less invasive than tumor biopsies and could enable improved monitoring and prediction of treatment outcomes for cancer patients.
The findings, published in the journal Nature, may support the development of personalized cancer therapies through enhanced treatment monitoring and outcome prediction.