CellsFromSpace: A versatile tool for spatial transcriptomic data analysis with reference-free deconvolution and guided cell type/activity annotation

Abstract

Spatial transcriptomics involves capturing the transcriptomic profiles of millions of cells within their spatial contexts, enabling the analysis of cell crosstalk in healthy and diseased organs. However, spatial transcriptomics also raises new computational challenges for analyzing multidimensional data associated with spatial coordinates.

Depending on the technology used, the capture of transcript is either at a near single-cell level or at a subcellular level. This introduces corresponding challenges, such as inferring cell composition in regions covering multiple cells or segmenting z-stacked cells from images to assign transcripts to specific cells.

In this context, we introduce a novel framework called CellsFromSpace. This framework allows users to analyze various commercially available technologies without relying on a single-cell reference dataset. Based on the independent component analysis, CellsFromSpace decomposes spatial transcriptomic data into components that represent distinct cell types or activities.

The framework provides a user-friendly graphical interface that enables non-bioinformaticians to perform a full analysis and to annotate the components based on marker genes and spatial distributions. The direct annotation of components, allows users to identify and isolate cell populations in the latent space, even when they overlap. Then, they can focus their studies on specific populations or analyze the proximity between different populations. Additionally, CellsFromSpace offers the capability to reduce noise or artifacts by component selection and supports analyses on multiple datasets simultaneously.

Here, we demonstrate the efficiency of CellsFromSpace to successfully identifies spatially distributed cells as well as rare diffuse cells on datasets from the Visium, Slide-seq, MERSCOPE, and COSMX technologies.