Microbiome intervention analysis using transfer functions and mirror statistics (Dr. Pratheepa Jeganathan)

Join us for a colloquium presented by Dr. Pratheepa Jeganathan! Refreshments provided.

This talk delves into the limitations of current models in studying microbiome interventions, particularly their lack of temporal memory and reliance on simplistic assumptions like Markovian dynamics and linearity. Addressing these shortcomings, we introduce a new class of models based on transfer functions. These models are designed to learn impulse responses, which are crucial in understanding the delayed effects that environmental changes can have on microbial communities. This methodology allows for the simulation of microbial ecosystem trajectories under hypothetical interventions, offering a way to identify significantly perturbed taxa while ensuring False Discovery Rate guarantees. 

The presentation will focus on the estimation procedure, the process of inference, and the utilization of counterfactual simulations in our research. Through a combination of simulations and real-world case studies, we will demonstrate the effectiveness and interpretability of these models. This is a joint work with Kris Sankaran. 

Dr. Jeganathan is an assistant professor in the Department of Mathematics and Statistics and an associate member of the School of Computational Science and Engineering at McMaster University. Their research focuses on challenges that arise in modern datasets and multi-view learning. Their methods development is motivated by applications in microbiome multi-omics, spatial multi-omics, sensor-based traffic data, and loss reserving in non-life insurance.

Their research is partially funded by NSERC. They are also one of the collaborators of the Collaborative Research Team's project funded by the CANSSI, the SEED grant project funded by The Office of the Provost and MacPherson Institute, and the Explorance Faculty Research Grant project on student evaluation of teaching.