Research

Integrating AI and Biology Across Scales

Our research group develops AI/ML methods to uncover the principles of cellular organization, communication, and function, with broad implications for health and disease. Despite remarkable progress in single-cell profiling and other high-throughput molecular and cellular assays, we still lack integrative models that connect molecular mechanisms to cellular behaviors and tissue dynamics. We bridge this gap by designing AI systems that integrate across molecular, cellular, and tissue levels to reveal how structure and communication give rise to function, with applications in critical areas such as neurodegeneration, cancer, and immunology.

Our vision is to build comprehensive multiscale models that elucidate how cellular systems operate - from chromosome folding in the nucleus to collective behaviors in tissues, and from normal development to disease processes. We develop interpretable and generalizable AI systems that not only analyze but also collaborate in discovery - proposing hypotheses, simulating perturbations, and guiding the next experiment. At the forefront of AI for biology, we focus on creating biologically informed models that enable a deeper mechanistic understanding of living systems, linking data, models, and experimentation toward the ultimate goal of decoding the language of cells to drive discovery in biology and medicine.

AI for Single-Cell 3D Epigenomics

We study how the genome folds within the nucleus and how its 3D organization influences gene regulation and cell identity. Our group has developed advanced AI/ML models that capture the spatial organization of chromatin, nuclear compartments, and associated molecular complexes by integrating data from imaging, Hi-C-like assays, and single-cell multi-omics. These frameworks enable in silico perturbations, hypothesis generation, and model-based reasoning about genome regulation.

AI for Spatial Tissue Dynamics

We develop AI models that integrate single-cell and spatial omics data to map how cells organize, interact, and communicate within tissues. Our methods model both intrinsic cellular states and their spatial dependencies, revealing how gene regulation and microenvironments shape tissue function. Ongoing work includes probabilistic modeling of tissue dynamics, self-supervised models of multiscale cellular interactions, and the development of cellular representations that generalize across tissues.

Generative AI and Biological Foundation Models

We build biological foundation models and generative AI systems trained on diverse data - from genomic sequences and single-cell profiles to imaging and structural maps. These models learn generalizable representations across scales and modalities, enabling transfer learning, mechanistic interpretation, and hypothesis generation. Our goal is to use generative modeling not only to describe but also to predict, simulate, and design new biological states, helping to reveal how molecular changes propagate to cellular and tissue phenotypes.

AI-Experiment Integration

We are developing AI agents that act as scientific collaborators - analyzing data, proposing hypotheses, and prioritizing experiments to maximize discovery. These models establish a continuous feedback loop between computation and experiment. By linking molecular perturbations, cellular phenotypes, and tissue outcomes, we aim to construct frameworks that not only predict but also guide the next generation of biological experiments.