BACT: nonparametric Bayesian cell typing for single-cell spatial transcriptomics data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics enables gene expression and spatial location measurement at cellular resolution.
• Existing methods often require pre-specifying cell type numbers, hindering exploratory analysis.
• Accurate cell type identification within tissue context is crucial for biological discovery.

Purpose of the Study:

• To develop a nonparametric Bayesian model for automated cell typing in spatial transcriptomics.
• To address the limitation of pre-specifying cell type numbers in existing clustering methods.
• To improve the accuracy and efficiency of spatial cell type analysis.

Main Methods:

• Developed BACT, a nonparametric Bayesian model integrating gene expression and spatial coordinates.
• Incorporated a nonparametric Potts prior to model spatial dependency between neighboring cells.
• The model automatically infers the number of cell types from the data.

Main Results:

• BACT demonstrated superior performance compared to existing spatial cell typing methods.
• Evaluations on three distinct spatial transcriptomic datasets validated the model's effectiveness.
• The method successfully identified cell types and their spatial distributions without prior number specification.

Conclusions:

• BACT offers a powerful, data-driven approach for cell type identification in spatial transcriptomics.
• The model overcomes key limitations of current methods, facilitating exploratory data analysis.
• BACT provides a valuable tool for uncovering fine-grained spatial patterns of cells within tissues.