We develop computational methods for multimodal biology, linking cell-state dynamics, tissue morphology, and spatial molecular measurements.
cellDancer
A deep-learning framework for RNA velocity with cell-specific kinetics. By estimating local velocities and relaying them across neighboring cells, cellDancer improves inference of multi-lineage cell-state transitions.
Nature Biotechnology, 2023Loki
A visual-omics platform built from OmiCLIP to align H&E histology with spatial transcriptomics. It turns paired image-omics data into a shared representation for annotation, retrieval, decomposition, and expression prediction.
Nature Methods, 2025Thor
A platform for cell-level spatial transcriptomics and histology. Thor uses anti-shrinking Markov diffusion to refine spot-level measurements toward cell-level tissue maps, paired with interactive whole-slide visualization through Mjolnir.
Nature Communications, 2025