Enhancing and accelerating cell type deconvolution of large-scale spatial transcriptomics slices with dual network...

Area of Science:

• Spatial transcriptomics
• Single-cell genomics
• Bioinformatics

Background:

• Cell type deconvolution integrates single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics data to map cell distributions within tissue slices.
• Existing methods often fail to leverage the full potential of both data types and are computationally intensive, limiting their use in large-scale studies.

Purpose of the Study:

• To introduce jMF2D, a novel joint learning nonnegative matrix factorization algorithm for efficient and accurate cell type deconvolution.
• To address the limitations of current algorithms by better integrating scRNA-seq and spatial transcriptomics data.

Main Methods:

• jMF2D employs a joint learning approach with network models to integrate scRNA-seq and spatial transcriptomics data.
• It jointly learns a cell type similarity network to improve cell type signature quality, enhancing deconvolution accuracy and efficiency.

Main Results:

• jMF2D demonstrates superior accuracy compared to state-of-the-art methods on diverse datasets.
• The algorithm achieves approximately 90% reduction in running time, making it suitable for large-scale analyses.
• jMF2D aids in identifying spatial domains and biomarker genes.

Conclusions:

• jMF2D provides an efficient and effective computational model for analyzing spatial transcriptomics data.
• The method enhances the integration of scRNA-seq and spatial transcriptomics, improving biological insights from tissue samples.