Our lab develops causal mediation frameworks for high-dimensional omics data, where classical causal assumptions are often strained or violated. Omics features are structured, correlated, compositional, and heterogeneous across scales—properties that challenge identifiability, ignorability, and stability assumptions underlying standard mediation analysis. We focus on adapting and extending causal methods to respect these biological structures rather than forcing them into simplified statistical models.

By integrating causal inference with modern high-dimensional modeling, our research aims to disentangle direct and indirect effects in complex biological systems while explicitly accounting for uncertainty, dependence, and model misspecification. This work provides principled tools for interpreting mechanistic pathways in genomics, epigenomics, and other omics domains, enabling more reliable causal conclusions from observational and longitudinal studies.