If you’re working with checkpoint inhibitors, you know these are powerful drugs for fighting cancer, but they don’t work for everyone. What if you could find reliable biomarkers to help you identify which patients would be most likely to benefit from this therapy?

A new model, called “HistoTME,” recently debuted by the Artificial Intelligence Research group in NCI’s Center for Cancer Research, could help pick the right patients for this type of treatment. Unlike traditional approaches, HistoTME learns to predict treatment responses by examining various properties of the tumor microenvironment (TME), a key factor for influencing responses to immunotherapy.

By applying the model to digital pathology images from patients with non-small cell lung cancer, the researchers were able to predict the tumor’s cell-type composition and track critical processes impacting tumor growth, thus boosting their ability to predict which patients would respond to treatment.

Dr. Sushant Patkar, first author on the recent study, noted, “Our model is a gamechanger for immune-oncology research as it offers a deeper understanding of the underlying molecular characteristics of the TME using routinely collected pathology slides. By applying our model, we could monitor changes in the tumor microenvironment more closely and in a cost-effective manner to select the best treatment options for patients.”

Dr. Baris Turkbey, a senior author on the study, said, “This study showcases how NCI is exploring innovative ways to incorporate AI into everyday clinical practices. Models such as HistoTME could significantly advance our understanding of the tumor microenvironment and its role in cancer treatment responses, leading to the discovery of new biomarkers for personalizing immunotherapy treatments.”