TITLE: An Introduction to Machine Learning Methods for Optimal Dynamic Treatment Decisions in Precision Medicine
SPEAKER: Prof. Donglin Zheng, UNC and Prof. Yuanjia Wang, Columbia University
Moderator: Kalyan Ghosh
This tutorial provides a comprehensive introduction to the development and application of machine learning methods for making optimal treatment decisions in personalized or precision medicine. In the first part, the tutorial begins by introducing causal inference under a potential outcome framework and then describes the concepts of heterogeneous treatment effects and dynamic treatment regimens. Next, the tutorial will illustrate how to use observational studies and sequentially randomized trials to estimate these quantities. Machine learning methods for estimating optimal dynamic treatment regimens or decisions will be introduced, with a focus on Q-learning and O-learning. The second part of the tutorial focuses on recent advancements in more complex settings, providing participants with a comprehensive overview of real-world applications. These advancements include the integration of real-world evidence (e.g., electronic health records), analysis of multi-domain outcomes, and consideration of the benefit-risk balance in decision-making.
Dr. Donglin Zeng is a professor at the Gillings School. Dr. Zeng is an expert on modern empirical process theory and semiparametric efficiency theory. He has made important contributions to survival analysis, causal inference, missing data, semiparametric models, statistical genetics, personalized medicine, machine learning and high-dimensional data.
He has published over 100 papers, most of which have appeared in first-tier statistical journals. Dr. Zeng is a Fellow of the American Statistical Association and a Fellow of the Institute of Mathematical Statistics.
Dr. Wang works on developing data-driven approaches to uncover complex relationships between biomarkers, clinical markers, environmental variables and health outcomes to assist studies of disease etiology, diagnostic capabilities and optimal treatments. Her methodological interests include machine learning, analytics for precision medicine, network analysis, and novel design and analysis of clinical trials and electronic health records. Her substantive applied research area includes psychiatric disorders and neurological disorders.