Research 


Our lab sits at the intersection of physics, mathematics, and cancer biology. We develop computational and theoretical frameworks to understand cancer as a dynamical process — not just a collection of mutations, but a transition between cellular states governed by gene regulatory networks.


1. Cancer Transition & Attractor Landscape Analysis


Cancer develops through a series of state transitions — from normal epithelium through precancerous lesions to invasive carcinoma. We model these transitions using attractor landscape theory from dynamical systems, where each stable cell state (normal, precancerous, cancerous) corresponds to an attractor in a high-dimensional state space.


By constructing Boolean network models and ODE-based dynamical models of gene regulatory networks, we identify the critical tipping points, bifurcation structures, and transition paths that drive tumorigenesis. This approach has revealed a percolation-like transition in colorectal cancer, where cooperative mutational effects accumulate until a critical threshold triggers irreversible transformation.


2. Control of Cancer Cell State Transitions


A central question in our lab: Can we identify molecular switches that control cancer cell fate? Through systematic attractor landscape analysis, we search for critical control nodes — specific molecular targets whose perturbation can redirect cellular state transitions, including driving cancer cells toward less malignant or normal-like states.


By mapping the full landscape of stable and unstable states within gene regulatory networks, we identify not only the transitions that lead to tumorigenesis, but also potential intervention points that could alter cell fate trajectories. We have computationally predicted and experimentally validated such control strategies in colorectal cancer models.


3. AI & Bioinformatics for Cancer Research


We develop and apply AI/machine learning methods tailored to cancer biology problems, spanning from molecular-scale modeling to patient-level prediction:



A major ongoing effort in our lab is the construction of patient-specific Digital Cell Twins for cancer. Using single-cell transcriptome data from colorectal cancer patients, we infer individualized Boolean network ensembles that capture the unique regulatory logic of each patient's tumor. These digital twins enable us to simulate drug responses, identify critical control targets, and predict cancer cell state transitions in silico — before any wet-lab experiment.


This work integrates single-cell SNV-based subclone identification, AI-driven dynamic GRN ensemble construction, and large-scale attractor landscape simulation to build a comprehensive computational framework for precision oncology.



We develop foundation model-based approaches for drug response prediction by leveraging large-scale single-cell transcriptomic data. By training on massive perturbation datasets encompassing diverse cancer cell lines and drug treatments, these models learn generalizable representations of how cells respond to pharmacological interventions. Our goal is to extend these cell-level predictions to more complex biological systems such as organoids and animal models through domain adaptation, ultimately bridging the gap between in vitro screening and clinical translation.


4. Gene Regulatory Network Dynamics


At the core of all our research is the study of gene regulatory network (GRN) dynamics. We investigate how feedback loops, bistable switches, and oscillatory circuits within GRNs give rise to complex cellular behaviors — from cell fate determination to uncontrolled proliferation.


Our earlier work in computational neuroscience, studying synchronization and phase-locking in neural circuits, provided the mathematical foundation that we now apply to cancer network dynamics. This cross-disciplinary perspective — rooted in physics and nonlinear dynamics — gives our lab a unique approach to understanding biological complexity.


Approach


Our research philosophy integrates three pillars: