Cross-modality mapping using image varifolds to align tissue-scale atlases to molecular-scale measures with...

Area of Science:

• Computational Biology
• Neuroscience
• Genomics

Background:

• Spatial transcriptomics technologies generate high-dimensional data, posing challenges for traditional atlas construction.
• Aligning sparse, emerging transcriptomic datasets with existing tissue atlases requires advanced computational approaches.
• Current methods struggle to model the complexity of subcellular gene detection and integrate diverse data types.

Purpose of the Study:

• To develop a computational framework for establishing correspondences between molecular-scale transcriptomics and tissue-scale atlases.
• To address challenges in atlas construction and cross-specimen/technology alignment with sparse data.
• To integrate diverse molecular and cellular datasets into a single coordinate system for comparison and further atlas development.

Main Methods:

• Representing spatial transcriptomics data as generalized functions encoding position and high-dimensional features (gene, cell type).
• Mapping data onto low-dimensional atlas ontologies by modeling regions as homogeneous random fields.
• Simultaneously solving for minimizing geodesic diffeomorphism (using LDDMM) and latent feature densities.

Main Results:

• Successfully mapped tissue-scale mouse brain atlases to MERFISH and BARseq transcriptomics data.
• Integrated gene-based and cell-based transcriptomics with histopathology and cross-species atlases.
• Demonstrated the utility of a unified coordinate system for diverse dataset integration.

Conclusions:

• The proposed method effectively bridges molecular-scale transcriptomics and tissue-scale atlases.
• This approach facilitates the integration of heterogeneous biological datasets for comprehensive atlas construction.
• Enables robust comparison and analysis across different technologies, species, and data modalities.