Abstract:
|
Recent advances in spatial barcoding-based transcriptomic technologies provide measurement of transcriptome-wide expression profiles and spatial locations of cells in tissue simultaneously. However, these technologies divide tissue into small capture spots and measure the average expression profile of a few unknown cells in each spot. Thus, the observed transcriptomic data lacks single-cell resolution. We consider the spatial correlation and the sparsity of cell types present in each spot and develop a graph Laplacian regularized model (GLRM) to deconvolve spatial barcoding-based transcriptomic data using reference scRNA-seq data from same tissue type with integration of spatial information. The model is efficiently optimized by an algorithm based on the alternating direction method of multipliers (ADMM). Our simulation and real data results suggest that by considering the physical proximity between spots in deconvolution of spatial transcriptomic data, GLRM greatly improved the accuracy of cell type composition estimation.
|