Skip to main content



Computational Immunohistochemistry

We perform deep learning of spatial immune signatures to aid in stromal immune cell identification. With the ability to provide ground truth to the spatial configuration of the tumor immune microenvironment, the ability for a deep learning algorithm to identify specific immune cell populations on routine H&E staining may help unravel novel immunologic mechanisms and better elucidate the role of immunologic biomarkers on specific immune cell populations. We focus on identifying these immune cell populations in human and murine tumors to allow for identification through computer vision techniques without the need for physical staining.


Predictive spatial immune signatures

Current immune biomarker development focuses on gross quantitation of biomarker levels on a slide, or single-cell techniques with an overabundance of data. Lost in these analyses is the spatial context of clinically relevant biomarker expression. We study the role of geospatial immune cell localization to better understand how immunologic structure may inform function.



We collect PBMC, sera, plasma, microbiome (stool, skin) from consented cancer patients on immunotherapy with clinical annotation to uncover novel immunobiology related to therapeutic response and toxicity