BASIN: Bayesian mAtrix variate normal model with Spatial and sparsIty priors in Non-negative deconvolution

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics provides gene expression data with spatial resolution.
• Current methods often lack cellular resolution, requiring cell type deconvolution.
• Inferring cellular composition is crucial for understanding tissue microenvironments.

Purpose of the Study:

• To introduce BASIN, a novel method for cell type deconvolution in spatial transcriptomics.
• To address limitations in cellular resolution of existing spatial transcriptomics data.
• To provide robust and accurate inference of cell type proportions.

Main Methods:

• Modeled deconvolution as a non-negative matrix factorization (NMF) problem with a graph Laplacian prior.
• Developed a matrix variate Bayesian NMF method incorporating nonnegativity and sparsity priors.
• Employed a Gibbs sampler to approximate posterior distributions and quantify uncertainty.

Main Results:

• BASIN outperforms existing deconvolution methods in accuracy and efficiency on diverse spatial transcriptomics datasets.
• The incorporated priors significantly influence the deconvolution results.
• The method provides a distribution of possible solutions, enhancing robustness.

Conclusions:

• BASIN offers a powerful and accurate approach for cell type deconvolution in spatial transcriptomics.
• The Bayesian framework provides inherent uncertainty quantification, crucial for biological interpretation.
• This method advances the analysis of spatially resolved gene expression data.