Research

My research connects gene relationships, spatial molecular programs, cell–cell interactions, and tissue dynamics through interpretable statistical and machine learning models.

01 Gene networks & cellular heterogeneity

I infer gene dependencies from single-cell counts while accounting for measurement noise and differences between cell populations.

Mixed round and elongated cell populations and single-cell counts linked to two population-specific undirected gene networks — Illustrative schematic · Cell heterogeneity and single-cell counts → cell-type-specific gene networks.

PLNet — gene network estimation from count data. Code
Insight: Count variation and latent gene dependencies should be modeled separately.

VMPLN — cell-type-specific networks in mixed populations. Code
Insight: Pooled cells can mix within-type gene relationships with differences between cell types.

02 Spatial & multimodal representations

I integrate complementary molecular and morphological measurements to connect interpretable biological programs with tissue architecture.

Expression measurements, H&E-style folded tissue morphology and nuclear chromatin linked to complementary molecular activity patterns across the same tissue anatomy — Illustrative schematic · H&E-style morphology and molecular measurements → spatial molecular programs.

SpaHDmap — high-resolution spatial metagenes from expression and histology. Code
Insight: Tissue morphology can inform molecular resolution while preserving interpretable spatial programs.

Alzheimer’s disease multiomics — a collaborative study of 3D genome organization and gene expression.
Insight: Joint measurements link chromatin reorganization to cell-type-specific disease alterations.

03 Cell communication & tissue niches

I model communication programs, neighborhood composition, and interactions across spatial scales to understand multicellular tissue organization.

A simple layered tissue context, teal and orange communication programs, and their matching spatial assignments to Niche 1 and Niche 2 — Illustrative schematic · Program 1 and Program 2 map to Niche 1 and Niche 2 using matching colors.

SpiderNet — an interpretable meta-interaction basis for cell–cell communication. Code
Insight: A shared basis separates communication programs from their activity across cell pairs.

SpaNiche — joint analysis of cellular colocalization and ligand–receptor patterns. Code
Insight: Neighborhood composition and interaction patterns together characterize tissue niches.

Steamboat — attention-based modeling of cellular interactions across spatial scales.
Insight: Cell-intrinsic programs, local communication, and long-range interactions contribute distinct information.

04 Tissue dynamics

This ongoing direction asks how molecular programs and cellular interactions change together across space and time, and how those changes relate to tissue responses in aging and disease.

Three tissue snapshots with a growing vessel-associated neighborhood, paired with separate molecular-program curves and changing cell-cell interaction networks — Illustrative schematic · Tissue remodeling over time, with program activity and cell–cell interactions shown separately.

Coordinated tissue change — modeling the temporal relationships between molecular programs, cell interactions, and tissue states.