Our research lies at the intersection of statistical physics, stochastic thermodynamics, computational biology, and information theory. We seek to understand how fundamental physical principles shape the behavior of living systems at the molecular scale. Life operates far from equilibrium, and processes such as gene expression and molecular regulation are governed not only by deterministic biochemical rules but also by stochastic variability and energetic costs. Using theoretical, computational, and stochastic modeling, we investigate how cells regulate gene expression, manage noise, allocate limited resources, and reliably transfer information. Thermodynamic principles provide a framework to quantify the energetic costs and constraints that govern these processes. To connect models with reality, we often integrate them with high throughput experimental datasets, including ribosome profiling, RNA sequencing, and proteomics, which link physical theory to measurable biological outcomes. Through this combined approach, we aim to uncover the design principles by which cells manage variability, maintain robustness, optimize resource use, and ensure faithful information transfer across molecular networks.
Stochastic Modeling of Gene Expression
Gene expression, the process by which the information encoded in DNA is converted into functional products, is one of the most fundamental activities of life. These products include both proteins, which carry out structural, enzymatic, and regulatory functions, and non-coding RNAs such as rRNAs, tRNAs, and regulatory RNAs, which are essential for protein synthesis and gene regulation.
In transcription, RNA polymerases initiate at promoter regions, elongate along the DNA template to synthesize RNA, and terminate once the transcript is complete. In translation, ribosomes decode messenger RNA into proteins through three stages: initiation, where the ribosome assembles at the start codon; elongation, where amino acids are sequentially added to the growing polypeptide chain; and termination, where the completed protein is released at a stop codon. The regulation of these processes determines the abundance, timing, and diversity of gene products within the cell.
The central outcome of gene expression is therefore the production of functional molecules whose levels and timing must be tightly controlled. This control is shaped by factors such as codon usage, elongation rates, mRNA structure, and the availability of ribosomes and polymerases. While these regulatory mechanisms set the average output, transcription and translation are inherently stochastic, with random binding and unbinding events, pauses, and collisions introducing variability. On top of this molecular randomness, cells employ feedback loops and global regulatory mechanisms—for example, proteins that repress or activate their own synthesis, or dynamic adjustments in ribosome availability that alter translation rates. These layers of regulation not only allow cells to maintain robustness, adapt to environmental signals, and optimize the use of limited resources, but also reflect an evolutionary dimension: codon usage patterns, initiation strengths, and elongation dynamics are shaped by natural selection to balance speed, accuracy, and efficiency, leading to conserved regulatory strategies across species.
Our research addresses these questions using a range of physical science–based approaches grounded in non-equilibrium statistical mechanics. We employ stochastic simulation methods, including Gillespie’s algorithm and Monte Carlo simulations, which act as virtual experiments by generating trajectories of molecular events that can be used to compute ensemble properties of gene expression. Complementing these numerical approaches, we formulate master equations to analytically describe the time evolution of system dynamics, and apply continuum approximations such as drift–diffusion equations and Fokker–Planck formalisms. These methods together provide a bridge between microscopic molecular interactions and system-level behaviour. We often integrate these approaches with high throughput datasets such as ribosome profiling and RNA sequencing to establish quantitative connections between theoretical predictions and experimentally observed patterns of gene expression.
Stochastic Thermodynamics of Living Systems
Living systems operate far from equilibrium, with many of their essential processes relying on continuous energy consumption. Energy expenditure allows biological processes to achieve levels of speed, precision, and control that equilibrium alone cannot support. A striking example is kinetic proofreading, which improves the accuracy of both transcription and translation by deliberately consuming extra energy to reject incorrect substrates, thereby enabling fidelity well beyond equilibrium limits.
We study these biological processes through the lens of stochastic thermodynamics, which provides a principled way to connect energy dissipation with performance. Entropy production quantifies the irreversibility of molecular events and reveals how energetic cost is linked to the achievable accuracy, efficiency, and speed of biological processes. This framing naturally leads to open problems at the heart of biology and physics: How much energy must be dissipated to reach a given level of efficiency? What thermodynamic trade-offs govern the strategies that living systems employ to regulate molecular processes and adapt to changing environments? And where, if at all, do physics-based boundaries emerge that limit what living systems can accomplish?
In this view, energy in biology is not only a regulatory resource but also an enabler of performance: it is invested to enhance efficiency, increase accuracy, accelerate reactions, and maintain robustness in the presence of molecular noise. Our research develops theoretical and computational approaches to uncover this energetic logic of life, aiming to map the trade-offs and boundaries that define how biological systems harness dissipation to achieve function
Data-Driven Biology and Bioinformatics
Advances in high-throughput technologies have transformed biology into a data-rich science. Techniques such as ribosome profiling, RNA sequencing, and quantitative proteomics generate massive datasets that capture cellular states at unprecedented resolution. These datasets contain signatures of regulatory processes and evolutionary pressures—but extracting such insights requires approaches that go beyond descriptive analysis.
Our research develops computational pipelines that analyze large-scale datasets while directly confronting them with theoretical models. The aim is not merely to catalog patterns, but to test whether models rooted in stochastic dynamics or thermodynamics can account for the observed trends. At the same time, the heterogeneity and noise inherent in large datasets demand careful interpretation: meaningful signals must be preserved while avoiding over-simplification. In this way, data analysis becomes a bridge between experiments and theory, ensuring that models are evaluated against reality while also being inspired by it.
This integration raises important questions: How can universal principles be distilled from complex datasets? Which features reflect evolutionary design, and which arise from biophysical constraints? And to what extent can large-scale data uncover organizing principles that remain invisible to either models or experiments alone?
Noise and Information Flow in Biological Systems
Biological systems function under conditions where randomness is unavoidable. At the molecular scale, binding and unbinding events, enzymatic reactions, signaling cascades, and gene regulation all occur stochastically, introducing noise into cellular behavior. This variability can propagate across scales—from fluctuations in transcription factor levels to differences in protein concentrations, metabolic fluxes, or even cell fate decisions. Yet, despite the pervasive presence of noise, biological systems manage to transfer information reliably, coordinating regulation and decision-making across multiple levels. Explaining how such reliability emerges in the face of randomness remains a central open problem at the interface of physics, biology, and information theory.
We study this problem by asking how noisy molecular interactions shape the ability of biological systems to convey signals and regulate responses. From this perspective, regulation can be viewed as an information problem: inputs such as environmental cues or regulatory signals must be translated into outputs such as gene expression patterns, metabolic states, or cellular decisions. Understanding the limits of this transfer is key to identifying the principles by which living systems achieve robustness.
Our research combines stochastic processes, statistical physics, and information theory to quantify these limits. Measures such as mutual information and channel capacity capture the fidelity of signaling in noisy pathways, while physics-based models reveal how molecular design and energetic costs influence information flow. In particular, we focus on the role of energy dissipation: how the consumption of energy allows biological systems to suppress noise and achieve reliable communication.This perspective motivates fundamental questions: How do living systems minimize the impact of noise while retaining adaptability? What strategies maximize information transfer under energetic and resource constraints? And what are the physics-based boundaries on reliable information flow in noisy molecular networks? By addressing these questions, our research seeks to uncover the principles by which life organizes information flow in the presence of stochasticity.
Computational Biology and Biophysics Lab 