Celloscope: a probabilistic model for marker-gene-driven cell type deconvolution in spatial transcriptomics data

Abstract

Spatial transcriptomics maps gene expression across tissues, posing the challenge of determining the spatial arrangement of different cell types. However, spatial transcriptomics spots contain multiple cells. Therefore, the observed signal comes from mixtures of cells of different types. Here, we propose an innovative probabilistic model, Celloscope, that utilizes established prior knowledge on marker genes for cell type deconvolution from spatial transcriptomics data. Celloscope outperformed other methods on simulated data, successfully indicated known brain structures and spatially distinguished between inhibitory and excitatory neuron types based in mouse brain tissue, and dissected large heterogeneity of immune infiltrate composition in prostate gland tissue.

Tumoroscope: a probabilistic model for mapping cancer clones in tumor tissues

“…Furthermore, studies combining H&E, WES, and ST for large cohorts of patients, could explore the dependencies between patient clinical features and the spatial patterns of clones found using Tumoroscope. Combined with cell-type deconvolution approaches for ST data in the tissue surrounding the tumors [39][40][41], our framework has the potential to bring unprecedented insights into the interactions of specific cancer clones, their phenotypes, and the surrounding microenvironment. In summary, Tumoroscope opens up a new avenue in cancer research with broad applications for a basic understanding of the disease and its clinical applications.…”

Joint cell type identification in spatial transcriptomics and single-cell RNA sequencing data

“…The three parameters: θ st , Z st and π st are defined in precisely the same manner as in Celloscope (6).…”

“…Nonetheless, the mini-bulk nature of ST poses the challenge of decomposing hidden cell-type mixtures within each spatial location (spot). To overcome this, we have developed Celloscope (6), specifically designed to infer the prevalence of each cell type from ST data. When scRNA-seq and ST data originate from the same tissue, an identical set of cell types can be assigned to individual cells in scRNA-seq or deconvoluted in each ST spot.…”

“…Nonetheless, the mini-bulk nature of some of popular ST protocols, such as (7) Visium (8) presents the challenge of decomposing hidden cell-type mixtures within each spatial location (spot). To tackle this issue, we have previously developed Celloscope (9), specifically designed to infer the prevalence of each cell type from ST data. Celloscope distinguishes itself from earlier methods for cell type deconvolution in ST spots (10)(11)(12)(13)(14)(15)(16)(17)(18)(19) by leveraging prior knowledge of marker genes for cell types.…”

Spatial transcriptomics deconvolution at single-cell resolution by Redeconve

“…The current algorithms can be categorized to two groups: (1) mapping-based methods, e.g., NovospaRc 1 , Tangram 2 , Celltrek 3 , and CytoSPACE 4 , which map single cells to the positions of ST data according to gene expression similarity or related measures; and (2) deconvolution-based methods, e.g., CARD 5 , RCTD 6 , cell2location 7 , DestVI 8 , SpatialDWLS 9 , SPOTlight 10 , STRIDE 11 , CellDART 12 , Celloscope 13 , DSTG 14 , and Stereoscope 15 , which try to reconstruct the ST observations by modeling the experimental process as sampling from different combinations of single cells. Mapping-based methods are superior to the current deconvolution-based methods regarding their single-cell resolution as the resolution of current deconvolution methods is limited to tens of cell types.…”