HydraNet: a Causal AI Model for Precision Immune Therapy Balancing Survival Benefits and Severe Immune Adverse Event Risk

Poster Number: P161
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Causal Inference, Real-World Evidence Generation, Deep Learning, Precision Medicine, Drug Discoveries, Repurposing, and Side-effect
Primary Track: Applications

Immune therapies have transformed cancer treatment, yet severe immune-related adverse events remain concerning. We developed HydraNet, a novel causal AI model, to elucidate the tradeoff between survival benefits and severe AKI risks across regimens involving immune checkpoint inhibitors. HydraNet leverages generalized propensity scores in the targeted regularization to enable causal inference for multiple treatments and enhance model stability. HydraNet performance is compared with multivariate regression models and inverse probability weights using N3C EMR data.

Speaker(s):
Yao Chen, Master of Science
Indiana University School of Medicine

Author(s):
Vithal Madhira - Palila Software; Xiaochun Li, PhD - Indiana University School of Medicine; Tyler Shugg, PharmD, PhD - Indiana University School of Medicine; Shadia Jalal, MD - Indiana University Melvin and Bren Simon Comprehensive Cancer Center; Fang-Chi Hsu, PhD - Wake Forest School of Medicine; Michael Eadon, MD - Indiana University School of Medicine; Benjamin Bates, MD - Department of Medicine, Rutgers-RWJMS Medical School; Noha Sharafeldin, MD, PhD, MSc - School of Medicine, University of Alabama at Birmingham, Birmingham; Umit Topaloglu, PhD - National Cancer Institute; Qianqian Song, Ph.D. - University of Florida; Jing Su, PhD - Indiana University School of Medicine;