Stereoscope posits a probabilistic model of spatial transcriptomics and an associated method for the deconvoluton of cell type profiles using a single-cell RNA sequencing reference dataset.

The advantages of Stereoscope are: (i) Can stratify cells into discrete cell types, (ii) Scalable to very large datasets (>1 million cells).

Stereoscope requires training two latent variable models (LVMs): one for the single-cell reference dataset and one for the spatial transcriptomics dataset, which incorporates the learned parameters of the single-cell reference LVM. The first LVM takes in as input a scRNA-seq gene expression matrix of UMI counts with cells and genes, along with a vector of cell type labels. Subsequently, the second LVM takes in the learned parameters of the first LVM, along with a spatial gene expression matrix with spots and genes. This matrix should be uploaded as a csv file with gene names, 3D location, and tissue values in different columns.

Example use case: Train a probabilistic model of spatial transcriptomics (measure all the gene activity in a tissue sample and map where the activity is occurring) and an associated method for the deconvoluton of cell type profiles (estimating the proportions of different cell types in samples collected from a tissue) using a single-cell RNA sequencing reference dataset.

Limitations: Effectively requires a GPU for fast inference.