STRIDE

STRIDE deconvolves cell-type composition in spatial transcriptomic data by leveraging topic profiles derived from single-cell transcriptomics to estimate cell-type proportions and spatial localization.

Key Features:

• Topic modeling from single-cell transcriptomics: Uses topic modeling techniques trained on single-cell transcriptomics profiles to generate topic models that represent cellular identities.
• Cell-type composition deconvolution: Decomposes spatially resolved transcriptomic spots into proportions of cell types using the derived topic profiles.
• Accurate estimation of cell-type proportions: Demonstrates high accuracy in estimating cell-type proportions with a balance of specificity and sensitivity.
• Mapping rare cell types: Identifies and maps rare cell types to specific spatial locations within tissue sections.
• Identification of spatially localized genes and domains: Detects genes and expression domains that are localized to specific tissue regions.
• Integration across successive sections for 3D reconstruction: Integrates deconvolution results across successive tissue sections to facilitate three-dimensional reconstruction of tissue architecture.
• Downstream analysis functions: Provides signature (topic) detection and visualization, spatial clustering based on neighborhood cell populations, and domain identification.

Scientific Applications:

• Developmental biology: Resolves spatially organized cell-type composition and gene expression patterns during tissue development by integrating single-cell and spatial transcriptomics.
• Oncology: Maps tumor and microenvironmental cell-type distributions and spatially localized genes to study heterogeneity and cellular interactions in cancer.
• Immunology: Locates immune cell subsets and their spatial niches within tissues to study immune organization and responses.

Methodology:

Applies topic modeling techniques (from text analysis) trained on single-cell transcriptomics to generate topic profiles and uses those profiles to deconvolve spatial transcriptomic data into cell-type proportions.