Uncovering the tissue molecular architecture at single-cell resolution could help better understand organisms’ biological and pathological processes. However, bulk RNA-seq can only measure gene expression in cell mixtures, without revealing the transcriptional heterogeneity and spatial patterns of single cells. Researchers at Zhejiang University have developed Bulk2Space, a deep learning framework-based spatial deconvolution algorithm that can simultaneously disclose the spatial and cellular heterogeneity of bulk RNA-seq data using existing single-cell and spatial transcriptomics references. The use of bulk transcriptomics to validate Bulk2Space unveils, in particular, the spatial variance of immune cells in different tumor regions, the molecular and spatial heterogeneity of tissues during inflammation-induced tumorigenesis, and spatial patterns of novel genes in different cell types. Moreover, Bulk2Space is utilized to perform spatial deconvolution analysis on bulk transcriptome data from two different mouse brain regions derived from our in-house developed sequencing approach termed Spatial-seq. Th researchers have not only reconstructed the hierarchical structure of the mouse isocortex but also further annotated cell types that were not identified by original methods in the mouse hypothalamus.
Workflow of Bulk2Space
a Overview of the design concept of Bulk2Space. Bulk transcriptome data is taken as the input, and a single-cell profile is used as the reference for characterizing the clustering space of the heterogeneous tissue. After deconvolution, input bulk data is deconvolved into single-cell transcriptomics data. Then, either of the two spatially resolved transcriptomics, spatial barcoding-based RNA-seq or image-based in-situ hybridization, is used as the spatial reference. Generated single cells are assigned to the corresponding coordinates based on the spatial reference. The output is a set of generated single-cell profiles with specified x and y spatial coordinates. b Detailed deconvolution procedure. The input vector of bulk tissue is equal to the production of the expression matrix of cell types and the proportion vector of each cell type. The calculated proportion of all cell types is employed for the subsequent single-cell generation. The single-cell reference is used to characterize the clustering space of the tissue, and a deep learning model generates single-cell profiles within the clustering space of each cell type. c, d demonstrate the strategies for spatial mapping based on the two mainly used spatially resolved transcriptomics approaches. c For spatial barcoding-based reference, each generated single cell is assigned to the spot with the highest gene expression correlation until the aggregation of cells within the spot is close enough to the exact expression value. d For image-based reference, each generated cell is assigned to the location where the cell on tissue has the highest similarity with the given cell.
Availability – https://github.com/ZJUFanLab/bulk2space