STsisal: a reference-free deconvolution pipeline for spatial transcriptomics data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics (ST) reveals tissue molecular states but lacks single-cell resolution.
• Existing reference-based deconvolution methods rely on scRNA data, which is often unavailable or incomplete.

Purpose of the Study:

• To introduce STsisal, a novel reference-free deconvolution method for ST data.
• To address the challenge of cell type identification in complex tissues without scRNA references.

Main Methods:

• STsisal adapts the SISAL algorithm for ratio matrix disentanglement.
• Integrates marker gene selection, mixing ratio decomposition, and cell type characteristic matrix analysis.
• Applies a reference-free approach to spatial transcriptomics data.

Main Results:

• STsisal precisely and efficiently discerns distinct cell types within complex tissues.
• Demonstrated superiority over existing deconvolution techniques through simulations and real-data application.
• Successfully unveiled intricate cell type composition in spatially resolved transcriptomic data.

Conclusions:

• STsisal provides a robust solution for cell type deconvolution in ST data.
• Offers a valuable tool for analyzing complex tissue microenvironments.
• Overcomes limitations of reference-based methods in spatial transcriptomics analysis.