Spatial biology
Mapping where molecular programs act within intact tissue.
Spatial transcriptomics measures gene expression while preserving the physical organization of a tissue, letting us ask not just which programs are active in a tumor but where they act and which cell populations they connect. Our SpaceMarkers framework identifies genes and molecular interactions associated with latent spatial features — pulling signaling and crosstalk between cell types directly out of spatial data without requiring predefined regions.
We apply these methods to spatial multiomic data from clinical samples — including pancreatic and hepatocellular cancers — to characterize how the tumor and its microenvironment are spatially organized, and how that organization shifts with treatment. Tooling: SpaceMarkers (R/Bioconductor) and the in-development TRACER for imaging-based platforms.
Single-cell analysis
Resolving tumor heterogeneity and immune response cell by cell.
Single-cell RNA sequencing resolves the mixture of cell states inside a tumor, revealing heterogeneity that bulk measurements average away. We build and apply computational methods to dissect this heterogeneity, characterize immune populations, and understand response and resistance to cancer immunotherapy.
This includes methods for inferring regulatory relationships from single-cell transcriptomics (for example, network inference using Granger-causality ensembles) and approaches that integrate single-cell data with spatial and imaging measurements so that cell-state information can be placed back into its tissue context.
Tumor microenvironment modeling
Mechanistic models that turn measurements into predictions.
Beyond describing the tumor microenvironment, we model it. Multi-scale and spatial quantitative systems pharmacology (QSP) models represent tumor–immune dynamics mechanistically, so that observations from one cohort can be turned into predictions — including virtual clinical trials that explore biomarker performance and treatment strategies in silico.
A central methodological theme is calibration: fitting these mechanistic models to multi-scale tissue data while quantifying the uncertainty in the resulting parameters and predictions, so the models remain honest about what the data can and cannot constrain.
Multi-scale integration & uncertainty
Connecting molecules, cells, and tissue into one coherent picture.
Cancer operates across scales — from molecular programs to individual cells to the architecture of whole tissues. Our overarching goal is to bridge these scales: connecting high-dimensional measurements to mechanistic understanding by integrating spatial, single-cell, and imaging omics with quantitative models.
This includes methods such as FunCN for defining functionally relevant cellular neighborhoods in spatial proteomics, and frameworks for propagating uncertainty through multi-scale analyses so that integrated conclusions carry well-characterized confidence.