For the first time, a blood test can reveal factors in tumor microenvironments across multiple cancer types, and could make precision oncology possible across a multitude of cancer patient populations, Mayo Clinic has announced.A new artificial intelligence framework, developed alongside Stanford Medicine researchers over the past eight years, offers a paradigm shift for clinical decision-making in oncology, according to Dr. Aadel Chaudhuri, a radiation oncologist at Mayo Clinic and professor in the Department of Radiation Oncology.Applying machine learning to liquid blood biopsies, which are non-invasive for patients, to profile spatially dependent cell states and multicellular ecosystems that tumors rely on, can help unprecedented insights that oncologists can use to better target cancer treatments.We spoke with Chaudhuri and Aaron Newman, associate professor of biomedical data science at Stanford, about the models behind their singular research, published in Nature earlier this month. Using AI to understand tumor microenvironments A major clinical question – how do cancer therapies impact tumor microenvironments – has motivated Chaudhuri and Newman for more than a decade"This has really been, in many ways, an open question clinically, because there are no assays to interrogate the tumor microenvironment over time," Newman said. "There's just nothing clinically."It's a question that had no answer until now, they told Healthcare IT News.They developed a deep learning model, or neural network, to understand how different components in TMEs organize into what they call "cancer neighborhoods," and could give doctors the unprecedented ability to predict how a patient will respond to a specific therapy – or resist it. Their focus is on epigenetics – examining the epigenome of normal cells surrounding a cancer. It's different from cancer genomics, which focuses exclusively on gene mutations within cancer cells.A tumor recruits and manipulates surrounding normal cells to help it survive. But these normal cells inside the TMEs do not have cancer mutations, meaning scientists are unable to use standard mutation tracking to find them.Chaudhuri and Newman developed AI to look for the tiny chemical tags attached to normal cell DNA, known as methylation markers, to find them. In addition to making groundbreaking discoveries about the mysterious TMEs, the AI framework can read these tags in a blood sample and separate the different signals. That allows medical scientists to pinpoint exactly what types of cells are active in a specific tumor's microenvironment.Cancer research has been highly focused on tumor genomics, Chaudhuri explained, when we asked if he knew of any other researchers working on TMEs (there aren't).Current FDA-approved liquid biopsies focus exclusively on circulating tumor DNA from malignant cells, and those biopsies are only possible for a small percentage of cancer patients. "We're just completely tunnel visioned into the tumor cells," he said.