A biopsy from a blood draw
A deep-learning model reconstructs which multicellular 'neighborhoods' make up a tumor — the kind of spatial map that used to require cutting tissue out — from cell-free DNA floating in a blood sample.
A solid tumor is not one thing. It is a patchwork of neighborhoods — immune cells, blood vessels, connective tissue and cancer cells arranged in recurring patterns that decide whether a drug works. Reading that architecture has always meant a biopsy: cut out a piece, slice it, image it. A Stanford-Mayo team just did it from a tube of blood instead.
Newman calls today's cancer therapy 'whack-a-mole,' and cancer cells 'like plants that thrive in some soils and die in others.' — Aaron Newman
The setup came first. Pooling more than ten million single cells and spatial-transcriptome spots across 132 tumor samples and ten cancer types, the group found that these neighborhoods aren't infinite — they collapse into nine recurring 'spatial ecotypes' that show up again and again across malignancies. Then the move that matters: dying tumor cells shed DNA into the bloodstream, and that DNA still carries methylation marks recording which genes were switched on. A neural network, Liquid EcoTyper, learned to read those chemical marks in plasma and infer which ecotypes a tumor is built from — no tissue required.
In 78 metastatic melanoma patients on immunotherapy, the blood readout tracked what a biopsy would have told them: two ecotypes flagged patients likely to get durable benefit, another flagged likely resistance, and it held in ten more patients from a different institution. Choosing who gets a checkpoint inhibitor is today closer to guesswork than most patients realize; a blood test that reads the tumor's terrain could turn it into a measurement.
The gap between the platform and the payoff is worth sitting with. Ten million cells and six imaging technologies went into defining the ecotypes — but the clinical claim rests on 78 melanoma patients, and the paper concedes its blood measurements 'need further validation across cancer types and clinical contexts.' One of the senior authors holds a financial stake in the company formed to sell it. The biology is real and the idea is elegant; whether it survives a prospective trial is the whole question.
The lenses
The facts
How this connects
Tap a node to open it