Leveraging Machine Learning and Real-World Data for Precision Oncology Care

James Zou, PhD

In an interview with guest expert James Zou, PhD, associate professor of Biomedical Data Science at Stanford University, learn about the potential for using machine learning and real-world data to identify mutations that predict patient responses to treatments for NSCLC and more. 

Please share a brief overview of your recent study. 

James Zou, PhD: We leveraged large real-world clinico-genomics data and machine learning to identify mutations that strongly predict how cancer patients respond to specific cancer treatments. 

What were your key findings? 

Dr Zou: We identified 776 somatic mutations associated with patient responses to immunotherapies, targeted therapies, and chemotherapies across 20 common types of cancer. 

How was the machine learning model developed to generate a risk score for response to immunotherapy in patients with advanced non-small cell lung cancer (aNSCLC) based on mutation profiles? 

Dr Zou: Using data of aNSCLC patients from Flatiron Health and Foundation Medicine, we trained a random survival forest model to generate a score for immunotherapy response. We showed that our score complements standard tumor mutation burden (TMB) scores. For example, patients who have low TMB but high scores by our measure tend to respond well to immunotherapies. 

How can leveraging large-scale real-world clinico-genomic data from electronic health records contribute to advancing precision oncology and improving patient outcomes in cancer care? 

Dr Zou: Our study demonstrates how we can now identify predictive biomarkers from real-world data. These markers are an important step in precision oncology because they inform treatment recommendations and shed light on mutation-treatment interactions.