CellsFromSpace: a fast, accurate, and reference-free tool to deconvolve and annotate spatially distributed omics...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics captures cellular gene expression within tissue context.
• Analyzing this multidimensional data presents significant computational challenges.
• Existing methods often require single-cell reference datasets.

Purpose of the Study:

• To introduce CellsFromSpace, a novel, reference-free analytical framework for spatial transcriptomics.
• To enable the analysis of diverse spatial transcriptomics technologies.
• To provide a user-friendly tool for cell type identification and data deconvolution.

Main Methods:

• Independent Component Analysis (ICA) is the core of the CellsFromSpace framework.
• The method decomposes spatial transcriptomics data into interpretable components.
• It does not require a single-cell reference dataset for analysis.

Main Results:

• CellsFromSpace successfully identifies spatially resolved and rare diffuse cell populations.
• Quantitative deconvolution was demonstrated across Visium, Slide-seq, MERSCOPE, and CosMX technologies.
• Comparative analysis showed CellsFromSpace to be faster, more scalable, and accurate than a reference-free alternative.

Conclusions:

• CellsFromSpace provides a flexible and powerful approach for spatial transcriptomics analysis.
• The framework is accessible to users without extensive bioinformatics expertise via a graphical interface.
• It facilitates noise reduction, subset analysis, and comprehensive downstream analysis of spatial data.