Cis-regulatory grammar
Learn how enhancer and promoter sequences determine regulatory activity.
Research
We turn coffee into code, code into algorithms, and algorithms into biological discoveries. ☕
Regulatory genomics
How does the genome encode cell-state-specific regulatory programs?
We develop computational models to understand how DNA sequence, chromatin accessibility, transcription factor binding, and regulatory elements control gene expression. Our goal is to move from descriptive maps of regulatory activity toward predictive models of regulatory function.
- DNA sequence
- Chromatin state
- Regulatory activity
- Gene expression
- Phenotype
Chromatin accessibility and TF binding
Infer transcription factor occupancy, footprints, and regulatory activity from chromatin measurements.
Regulatory element discovery
Identify enhancers, promoters, and functional non-coding elements from genomic and epigenomic data.
Variant-to-function modeling
Predict how genetic variation alters regulatory activity and downstream gene expression.
Multimodal cellular modeling
How can we model cell state from sparse, multimodal measurements?
Single-cell and spatial assays capture different layers of cell state, but their measurements are noisy and incomplete. We develop probabilistic and machine-learning methods to integrate these layers and reconstruct cellular states, regulatory programs, and tissue organization.
- Molecular measurements
- Integrated cell-state representation
- Tissue context
Cell state connects genome regulation, the transcriptome, the epigenome, spatial context, and perturbation response.
Single-cell multi-omics
Jointly model chromatin accessibility, gene expression, and other molecular modalities.
Cell-state representation
Learn biologically meaningful representations of cells for comparison, retrieval, and prediction.
Development and perturbation
Model transitions between cell states during differentiation, disease, and experimental perturbation.
Spatial organization
Combine spatial context and molecular profiles to understand tissue architecture and cell–cell interactions.
Predictive models of cellular systems
Can we predict how cells respond to genetic and environmental perturbations?
Our research explores machine-learning and generative models that learn reusable representations of cellular systems. We aim to move beyond analyzing observed cell states toward predicting perturbation responses and guiding experimental design, with predictions tested alongside experimental collaborators.
- Observation
- Representation
- Prediction
- Perturbation
- Experimental validation
Perturbation modeling
Explore models of transcriptional and regulatory responses to unseen genetic or molecular perturbations.
Biological foundation models
Investigate how pretrained models can be adapted and evaluated for single-cell and regulatory genomics.
Cell search and retrieval
Explore shared representations that connect cells, molecular states, phenotypes, and biological knowledge.
Experimental design
Use predictive models to prioritize informative perturbations, measurements, and follow-up experiments.
Biological and disease applications
We work with experimental and clinical collaborators on myocardial infarction, fibrosis, systemic sclerosis, kidney biology, diabetes, and immune regulation. These biological questions inform method development and provide settings in which to test models of genome regulation and cell state.
Myocardial infarction
Map cellular responses and spatial tissue remodeling after heart injury.
Fibrosis
Investigate the cell states and regulatory programs associated with tissue scarring.
Systemic sclerosis
Study fibroblast–macrophage interactions and their spatial organization in affected tissues.
Kidney biology
Explore kidney cell identity, injury responses, and organoid models of disease.