Integration of spatial and single-cell data across modalities with weak linkage.

Area of Science:

• Computational biology
• Genomics
• Proteomics

Background:

• Single-cell sequencing and spatial omics technologies provide high-resolution molecular data.
• Integrating these multi-modal datasets is crucial for a comprehensive understanding of biological systems.
• Current cross-modal integration methods struggle with datasets lacking strong, pre-defined feature linkages, a scenario termed 'weak linkage'.

Approach:

• Developed MaxFuse, a novel computational method for cross-modal data integration.
• MaxFuse employs iterative co-embedding, data smoothing, and cell matching to leverage all available information.
• The method is modality-agnostic, allowing flexible integration of various omics data types.

Key Points:

• MaxFuse demonstrates high robustness and accuracy, particularly in weak linkage scenarios.
• Benchmarks on multi-omic datasets confirm the method's effectiveness.
• Successfully integrated spatial proteomic data with single-cell sequencing data, consolidating information at single-cell resolution.

Conclusions:

• MaxFuse overcomes limitations of existing methods in cross-modal omics data integration.
• Enables spatial consolidation of transcriptomic, epigenomic, and proteomic information within tissue sections.
• Offers a powerful tool for advancing multi-modal single-cell and spatial omics research.