Research

My research sits at the intersection of machine learning and tissue biology, with an emphasis on making spatial and imaging technologies more accurate, interpretable, and usable. Below is a more detailed look at the directions I work on; see Selected Works for the publications behind each.

Spatial transcriptomics

Computational analysis and benchmarking of imaging-based (e.g., Xenium) and sequencing-based (e.g., Visium) spatial transcriptomics data, with a focus on understanding how technology choice shapes the resulting data and downstream biological interpretation.

Digital pathology & deep learning

Predicting spatial gene expression directly from hematoxylin and eosin (H&E) histology images using deep learning, and studying how training data quality — molecular sparsity/noise and image resolution — affects model performance and interpretability.

Data & probe quality

Assessing measurement fidelity in spatial transcriptomics platforms, including off-target probe binding in probe-based panels, to improve the biological interpretability and reproducibility of spatial data.

Subtype discovery

Feature-selection and metric-learning methods that preserve heterogeneity while distinguishing known disease or cell states, in order to uncover disease and cell-state subtypes from transcriptomic and proteomic data.

Live-cell image analysis

Deep learning for cell segmentation, tracking, and morphodynamic profiling in live-cell microscopy, including work on phase-contrast and fluorescence imaging pipelines.

Open-source software

Building user-friendly, reproducible computational tools for biologists who may not have a strong programming background — see the Software page for released tools.