AI Model Developed to Predict Cancer Immunotherapy Response from Genetic Data

Researchers at Pohang University of Science and Technology in South Korea have developed an artificial intelligence method that analyzes cancer patients' genetic information to predict their response to immunotherapy. The findings were published in the journal Digital Medicine.

Immunotherapy works by activating a patient's own immune system to fight cancer, offering longer-lasting benefits in some cases, but not all patients respond. Identifying likely responders in advance could help avoid unnecessary costs and delays in pursuing alternative treatments.

The tumor microenvironment varies greatly between individuals, and senescent cells can suppress immune activity. Limited patient samples and vast genetic data also pose challenges for AI modeling. To address this, the team adopted a stepwise approach: first identifying tumors with a senescent microenvironment, then stratifying patients by their likelihood of responding to immunotherapy. This method aims to reduce model bias and improve predictive stability across different cohorts.

In an independent skin cancer patient cohort not included in model training, the method maintained consistent predictive performance. The team said the study incorporates the previously overlooked tumor senescent microenvironment into the prediction framework, potentially aiding more personalized treatment decisions.