X

AI-driven spatial omics

About the Project

Humans are composed of roughly 37 trillion cells, each with distinct biological functions and molecular profiles. While single-cell sequencing has provided valuable insights, spatial omics technologies now enable measurement of molecular information within the original tissue context, revealing how cells are organized and communicate. Platforms such as Visium HD, Xenium, MERSCOPE, CosMx, and PhenoCycler provide complementary strengths in terms of resolution and molecular coverage, generating high-content maps of tissues at single-cell or even subcellular scale.

In parallel, histology imaging of the same tissue sections offers detailed structural information at subcellular resolution. These images provide a rich morphological backdrop that complements spatial omics data, allowing us to see how molecular patterns align with tissue architecture.

Our lab focuses on AI-driven spatial omics, developing machine learning approaches that integrate histology and molecular data. This paradigm enhances resolution, scalability, and interpretability, enabling deeper insights into tissue biology and disease mechanisms. We have developed a series of tools that aim to combine the strengths of spatial omics and pathology imaging so that we can achieve scalability, accessibility, and speed.

  • SpaGCN: Integrating spatial location, gene expression, and histology for spatial domain detection (Hu et al. 2021 Nature Methods)
  • TESLA: Deciphering tumor ecosystems at super resolution from spatial transcriptomcis (Hu et al. 2023 Cell Systems)
  • SpaDecon: Cell-type deconvolution in spatial transcriptomics with semi-supervised learning (Coleman et al. 2023 Communications Biology)
  • CeLEry: Leveraging spatial transcriptomics to recover cell locations in single-cell RNA-seq (Zhang et al. 2023Nature Communications)
  • iStar: Enhancing spatial resolution of gene expression in platforms that lack single-cell resolution (Zhang et al.2024 Nature Biotechnology)
  • MISO: Multimodal spatial omics modeling (Coleman et al. 2025 Nature Methods)
  • iSCALE: Scaling up spatial transcriptomics for large-sized tissues that are beyond conventional platforms (Schroeder et al. 2025 Nature Methods)
  • S2Omics: Designing smart spatial omics experiments (Yuan et al. 2025 Nature Cell Biology)

Category :

Statistical and Computational Methods