Ongoing Research
Multimodel inference to account for model uncertainty in systems biology
How do we reconcile having multiple models of the same biological process? How does having multiple models effect the quality of our predictions?
I explore these questions using statistical methods for multimodel inference, model averaging, and model fusion in this work. I am currently focusing on applications to cell signaling models.
Uncertainty quantification for predictive modeling in diabetes
Mathematical modeling plays an important role in diagnosing diabetes and predicting pancreatic function from available clinical data. I am working to understand how UQ can improve predictive modeling in this setting.
Data-Driven patient phenotyping in diabetes
In collaboration with researchers from the UCSD School of Medicine, I am working to understand the progression of complications related to diabetes. We are taking a data-driven approach, specifically analyzing clinical records to predict different phenotypes that align with a propensity to develop a complication.
Completed Projects
Bayesian Model Calibration for Systems Biology
In this work, we developed a framework for Bayesian parameter estimation for dynamical models in systems biology that combined identifiability analysis, sensitivity analysis, and Bayesian inference.
Bayesian inference for uncertainty quantification in composite materials failure analysis
Collaboration with materials science researchers Johannes Reiner (Deakin Univ), Navid Zobeiry (UW, Seattle), and Reza Vaziri (UBC) and my advisor Boris Kramer (UCSD).
We utilized neural network surrogate models to enable Bayesian parameter estimation of composite material properties in finite element simulations for failure analysis.
Visualizing structure in widefield optical calcium recordings of the developing mouse cortex: FLOW Portraits
Working with Bing Brunton (UW Biology), Steve Brunton (UW Mech. Eng), William Moody (UW Biology), Dennis Tabuena (UW Neuroscience), and Nicholas Steinmetz (UW Neuroscience), I developed a novel method to analyze optical recordings of cortical neural activity.
The new method, adapted techniques from experimental fluid mechanics to visualize structure in the spatiotemporal neural recordings data.