In October 2025, Rick Chang, a PhD student in the Department of Biostatistics and Health Data Science, published his first-author paper “High-dimensional causal mediation analysis by partial sum statistic and sample splitting strategy in imaging genetics application” in Bioinformatics. The work represents a significant step forward in understanding how genetics and the environment jointly influence disease through complex biological pathways.
Causal mediation analysis aims to understand how and why an exposure, such as a gene or environmental factor, affects an outcome, often through intermediate variables known as mediators. But when those mediators come from high-dimensional sources like imaging or omics data, traditional approaches struggle. Chang and his team, under the mentorship of Pitt Public Health Professor George Tseng and Kayhan Batmanghelich, took on this challenge by developing a new statistical framework called the Partial Sum Statistic and Sample Splitting Strategy (PS5).
The PS5 framework combines rigorous theoretical design with practical efficiency. It helps researchers handle high-dimensional mediators, like thousands of imaging voxels, while maintaining interpretability and statistical accuracy. Through simulations, Chang demonstrated that PS5 controls type I error, increases statistical power, reduces bias, and more precisely identifies true mediators compared to existing methods.
Applying PS5 to imaging genetics data from the COPDGene study, Chang and his collaborators uncovered key insights into chronic obstructive pulmonary disease (COPD). Their analysis revealed that CT imaging mediates nearly 50% of the genetic effect on lung function, overlapping significantly with the effects of smoking. They also pinpointed a region in the lower lobe of the lung that plays a central role in mediating both genetic and environmental risks, a discovery that could inform future treatment strategies for mitigating COPD severity.
Chang credits the mentorship of Tseng and Batmanghelich, faculty at Boston University, and the collaboration of former students Yusi Fang and Michael T. Gorczyca for shaping the research. “I was inspired by the challenge of translating statistical innovation into meaningful biomedical discovery,” he explains. “My goal was to bridge rigorous causal inference theory with real-world biological complexity.”
-Calvin Dziewulski