deconvolution-cell2location

What problem does it solve?

Reference-based Bayesian deconvolution of spatial transcriptomics to estimate cell-type proportions per spot, providing uncertainty estimates.

Core Features & Use Cases

• Two-stage model: learn cell type signatures from scRNA-seq reference, then map to spatial spots.
• Uncertainty quantification: provides credible intervals for each cell type proportion.
• Output integration: stores results in adata.obsm['deconv_weights'] and adata.obs['celltype'] for downstream analysis.
• Use Case: Deconvolve Visium data to map cell-type composition across tissue sections.

Quick Start

Load a annotated scRNA-seq reference and a Visium spot dataset to run Cell2location and obtain per-spot cell-type proportions with uncertainties.