Thor

Thor integrates spatial transcriptomics and high-resolution histological images to infer single-cell spatial transcriptomes and relate gene expression to tissue morphology.

Key Features:

• Anti-Shrinking Markov Diffusion Method: Employs the Anti-Shrinking Markov Diffusion Method to infer single-cell spatial transcriptomes from spot-level data and integrate cell morphology with spatial transcriptomics.
• 10 analysis modules: Provides 10 modules tailored for genomic and image analysis at cellular resolution.
• Mjolnir integration: Integrates with Mjolnir for gigapixel histology image handling and linking histology to gene expression, pathway enrichment, and immune response.
• Scalable spatial workflows: Includes scalable workflows for spatial feature extraction, cell segmentation, gene–spatial correlation, and visualization.
• Semi-supervised fibrotic annotation: Supports semi-supervised annotation of fibrotic regions across heart failure zones.
• High-resolution data imputation: Performs high-resolution spatial transcriptomics data imputation demonstrated on Visium HD.
• Unbiased hallmark screening: Enables unbiased screening of breast cancer hallmarks.

Scientific Applications:

• Validation: Accuracy validated through simulations and datasets including ISH, MERFISH, Xenium, and Stereo-seq.
• Joint genomic–histology analysis: Performs joint genomic–histology analysis across diverse spatial transcriptomics datasets.
• Clinical insights in heart failure: Identified a regenerative signature in heart failure patient samples with protein presence confirmed by immunofluorescence staining.
• Structural elucidation: Revealed the layered structure of the mouse olfactory bulb and enabled unbiased screening of breast cancer hallmarks.
• Immune response heterogeneity: Elucidated heterogeneity in immune responses across different tissue types.
• Fibrotic region annotation: Annotated fibrotic regions within multiple heart failure zones using a semi-supervised approach.
• High-resolution imputation in Visium HD: Imputed high-resolution spatial transcriptomics data in a Visium HD bladder cancer sample, uncovering spatial patterns that align with histology.

Methodology:

Uses the Anti-Shrinking Markov Diffusion Method and scalable workflows for spatial feature extraction, cell segmentation, gene–spatial correlation, and visualization; includes semi-supervised fibrotic region annotation and high-resolution spatial transcriptomics imputation.