Our research

Mathematical foundations.
Biological understanding.

We develop theory, mechanistic models, and data-driven methods to understand how complex systems evolve and how macroscopic behavior emerges from underlying stochastic processes.

Our research spans probability and statistical mechanics, cellular dynamics, human physiology, and treatment response. A common goal is to connect interpretable mathematical structure with quantitative predictions.

01 · Inference from snapshots

Reconstructing continuous cellular dynamics

How can we recover continuous changes when individual cells cannot be followed over time?

Single-cell sequencing measures cellular states in high dimensions, but generally destroys the measured cells. We develop approaches using optimal transport, stochastic dynamics, gradient flows, and machine learning to reconstruct continuous dynamics from population-level snapshots.

Heterogeneous paths and cell fates

Our optimal-transport study of epithelial-to-mesenchymal transition (EMT) identified diverse transition paths and phenotypic outcomes among cells exposed to the same stimulus. This work connected mathematical modeling and computational analysis with EMT biology, in collaboration with Herbert Levine’s group at Northeastern University.

PROFET

PROFET combines particle-based gradient-flow modeling with neural force matching to infer a continuous, time-dependent vector field from discrete scRNA-seq measurements. The framework reconstructs nonlinear dynamics and heterogeneous cellular responses across biological systems.

The work brought together gradient-flow theory from Markos Katsoulakis and collaborators at the University of Massachusetts Amherst with cell-line data from Rinath Jeselsohn’s group at Dana-Farber Cancer Institute.

Selected work: PNAS (2024) · Cell Systems (2026)

Directions ahead

We aim to extend these approaches to coupled cell-state and population dynamics, spatial and multimodal measurements, and dynamic regulatory networks, with potential applications in differentiation, development, and regeneration.

02 · Coupled systems

Mechanistic modeling of biology and physiology

How do interacting processes across different timescales shape biological responses?

We build mechanistic models that connect biological state transitions, time-dependent external inputs, and interacting processes. Our approach combines mathematical formulation, computational implementation, parameter calibration, and model testing.

Biological and pharmacological dynamics

In our work on estrogen receptor-positive breast cancer, we coupled cell-cycle dynamics with drug pharmacokinetics and pharmacodynamics. This framework enabled quantitative predictions and systematic comparisons of palbociclib–fulvestrant treatment schedules.

Selected work: Cancer Research Communications (2023) · Publication details →

Human physiology across scales

Our broader research direction is to develop dynamical systems models of interacting cellular, tissue, and systemic processes. These models help formulate biological hypotheses, interpret measurements, and explore how interventions may change system behavior.

03 · Randomness and dynamics

Stochastic processes in cell biology and physiology

How do fluctuations shape the behavior of nonlinear systems?

We study the relationship between underlying stochastic processes and observable dynamical behavior. Our mathematical background includes stochastic limit-cycle oscillations and nonlinear systems under random perturbations.

This perspective motivates questions about variability, oscillatory behavior, and the emergence of reproducible patterns in cell biology and physiology.

Selected work: Stochastic Limit-Cycle Oscillations of a Nonlinear System Under Random Perturbations · Journal of Statistical Physics (2021) · Publication details →

04 · Probability and statistical mechanics

From stochastic systems to macroscopic laws

Why do canonical exponential distributions emerge when a subsystem interacts with a much larger environment?

Our work on the probabilistic foundations of the Gibbs canonical ensemble established a conditional limit law showing how canonical-type exponential distributions emerge as the surrounding system becomes large, without assuming the canonical distribution in advance.

Developed during Yu-Chen’s PhD under Hong Qian’s supervision, with Yizhe Chu contributing to the mathematical theory and proofs, this work connects statistical mechanics with general asymptotic probability theory.

Its applications lie primarily in probability and statistical mechanics. It also informs a guiding principle of our research: interpretable macroscopic laws can emerge from underlying stochastic systems through identifiable mathematical limits.

Selected work: Asymptotic Behavior of a Sequence of Conditional Probability Distributions and the Canonical Ensemble · Annales Henri Poincaré (2021) · Publication details →