Research Directions

Our group develops Computational, AI & Theory-driven approaches to understand how Biomolecules self-organize, communicate, and compute across scales—from single proteins to genome-wide networks. We combine physics-grounded modeling with modern machine learning to tackle fundamental questions in molecular and cellular biophysics.

Allostery & Biomolecular Signaling

Allosteric regulation, where a signal at one site controls function at a distant site, is fundamental to nearly every cellular process. We develop AI- and physics-based methods, including chemically accurate contact response analysis (CHACRA) and coevolutionary approaches, to map allosteric pathways in proteins and multi-protein assemblies. Applications range from enzyme regulation and thermoadaptation to understanding how mutations dysregulate signaling in disease.

Key methods: Coevolutionary analysis, contact response analysis, AlphaFold-guided coarse-grained simulations, enhanced sampling.

Biomolecular Condensates & Phase Separation

Liquid-liquid phase separation organizes the cell interior into membraneless organelles, hubs for gene regulation, stress response, and RNA processing. We use multi-scale models spanning atomistic to field-theoretic descriptions to decode how protein and RNA sequence encodes condensate material properties, viscoelasticity, and aging. We collaborate closely with experimental groups to connect molecular grammar to observed condensate behavior.

Key methods: Coarse-grained simulations, field-theoretic models, energy landscape theory, machine learning analysis of condensate dynamics.

Chromatin Organization & Epigenetics

The three-dimensional organization of chromatin inside the nucleus governs gene expression programs. We build mesoscopic liquid models of the nucleus (MELON) that capture how compartmentalization, lamina interactions, and transcription factor dynamics shape nuclear architecture. These models connect histone modifications and epigenetic states to genome-scale functional outcomes.

Key methods: Mesoscale liquid models, polymer physics, 4D simulations of nuclear organization, systems biology.

Generative & Explainable AI for Biomolecules

Machine learning is transforming structural biology, but interpreting what models have learned remains a challenge. We develop generative AI frameworks, including structure-aware protein sequence models and deep unsupervised learning tools, that not only make predictions but provide physically interpretable insights into sequence-structure-function relationships.

Key methods: Graph neural networks, variational autoencoders, structure-aware generative models, explainable AI.