UW Bio UW ChemE eScience
Lab Affiliations: NU CSB UW CMB

The Bagheri Lab integrates experimental data with computational strategies to uncover fundamental principles governing intracellular dynamics and intercellular regulation. Our group is highly collaborative, uniting a diverse range of research interests to tackle grand challenges from predicting and controlling the complex dynamics of cell populations to developing new methods for prediction and analysis. A unifying thread across all of our projects is the drive to elucidate, predict, and ultimately control biological response, particularly in the context of disease. When the regulation of biological function breaks down, the consequences can manifest as a range of illnesses, including cancer.

Our collaborative projects span four major thematic areas. Emergent dynamics examines how complex, higher-level population behaviors arise from lower-level interactions within and among cells. Deep learning applies neural network-based algorithms to extract patterns from biological data for prediction and analysis. Control theory employs computational frameworks to design targeted interventions that steer biological systems toward desired states, with direct implications for therapeutic strategy. Network theory represents biological systems as graphs to represent and study the structure of cellular interactions.

Effective prediction of cell population responses to intrinsic perturbations and extrinsic interventions requires the integration of intracellular signaling with intercellular dynamics. Using spatiotemporal models, we simulate the behavior of heterogeneous cell populations and trace how subcellular changes propagate into diverse population-level responses.

Simulation of tumor angiogenesis

Krista Phommatha

emergent dynamics

Vasculature is vital for tumor growth as it supplies tumors with the necessary oxygen and nutrients for cancer cell growth. As tumors grow in size, angiogenesis occurs, where hypoxic conditions in the TME recruite surrounding vasculature to grow capillaries. We have developed a model of vasculature in ARCADE, and are extending this model to include angiogenesis. We are using this angiogenic vasculature model to interrogate how perturbations in angiogenesis can be used to mitigate tumor growth.

Combinatorial antigen cell therapy design

Allison W. Li

emergent dynamics

Chimeric Antigen Receptor (CAR) T cell therapy has demonstrated the power of designing endogenous T cells as a tool to target hematologic tumors. However, solid tumors introduce unique challenges due to diverse antigen presentation as well as the evolving solid tumor microenvironment (e.g. oxygen tension, immune privileged regions, angiogenesis). We build in silico models of cell-based therapies that combinatorially target multiple antigens in the context of a dynamic tumor microenvironment to broaden the viable targets for these therapeutic interventions.

Follicular wave dynamics under hormonal perturbations

Krista Phommatha

emergent dynamics

The complex interplay between ovarian follicle recruitment and hormone dynamics greatly impact assisted reproductive therapies and fertility preservation protocols. An emergent phenomena observed in ovarian follicle recruitment is how and when they are recruited in groups around similar times. We are developing a mathematical model of ovarian follicle development under hormonal regulation to investigate how the hormonal perturbations that occur during fertility preservation treatments effect follicle recruitment dynamics.

Spatial structure and eco-evolutionary stability of microbial consortia

Madeline Scott

emergent dynamics

Engineered multispecies microbial communities have the potential to expand the functional capabilities of engineered microbes in medicine and bioproduction beyond what a single engineered strain can achieve. A key challenge is achieving ecological stability, and one promising approach is to draw from natural systems, where microbes often exhibit costly obligate mutualism—interactions characterized by interdependence and active cooperation. Here, we use a multiscale agent-based model, SEED, to investigate the eco-evolutionary processes underlying the emergence of these stable partnerships. We specifically examine how environmental factors, such as spatial structure and resource influx, can be modulated to promote trajectories toward obligate mutualism.

Collaborators: B. Kerr

Asymmetric cell division in Drosophila neurogenesis

Sophia K. Jannetty

emergent dynamicsnetwork theory

Neurogenesis is the process by which neurons are generated from neural stem cells. In Drosophila, neurogenesis requires a series of asymmetric cell divisions that yield daughter cells with distinct sizes and identities. The mechanisms determining the intracellular locations of these cell divisions and how these distinct identities are established are not yet fully understood. We are developing a spatiotemporal model to explore the dynamics of cell division and differentiation during Drosophila neurogenesis, with the goal of identifying the key factors that regulate where cell division occurs and how daughter cell identities are determined.

Collaborators: C. Cabernard

Wound healing in aging tissue

Sophia K. Jannetty

emergent dynamics

Wound healing requires coordinated migration and proliferation of several cell types. With age, senescent cells accumulate in tissues and secrete inflammatory factors that disrupt this coordination, contributing to delayed closure, chronic inflammation, and fibrosis. How cellular senescence shapes repair at the tissue scale is not well understood. We are developing a spatiotemporal model of wound healing that incorporates age-related changes in vasculature, immune clearance, and senescent load to identify when senescence promotes versus impairs repair, and to evaluate therapies that could restore healthy healing in aging tissue.

Spatial control of agent-based models

Jacob I. Evarts

deep learningcontrol theory

Controlling agent-based models (ABMs) is challenging due to high-dimensionality and frequently stochastic outputs. While lower-dimensional ODE surrogate models can be used to simplify these systems, they aggregate data in a way that discards spatial information. We are developing a deep learning pipeline to encode ABM outputs into lower dimensional spaces that capture underlying dynamics while maintaining spatial context, enabling precise, spatially-constrained control interventions.

Histological validation of spatiotemporal tumor models

Po-Hao Chiu

deep learning

Spatiotemporal models can reproduce diverse tumor microenvironments, but comparing simulated spatial organization directly to patient tissue remains difficult. This project bridges simulation and experiment by rendering ARCADE agent-based simulations into H&E images with a conditional generative model, then comparing generated and real tissue within distinct tissue archetypes, such as tumor epithelium (top) and vascularized stroma (bottom). This tests whether the model reproduces each histological pattern without relying on human-curated metrics.