Accurate and Flexible Single Cell to Spatial Transcriptome Mapping with Celloc

Abstract

Accurate mapping between single-cell RNA sequencing (scRNA-seq) and low-resolution spatial transcriptomics (ST) data compensates for both limited resolution of ST data and missing spatial information of scRNA-seq. Celloc, a method developed for this purpose, incorporates a graph attention autoencoder and comprehensive loss functions to facilitate flexible single cell-to-spot mapping. This enables either the dissection of cell composition within each spot or the assignment of spatial locations for every cell in scRNA-seq data. Celloc's performance is benchmarked on simulated ST data, demonstrating superior accuracy and robustness compared to state-of-the-art methods. Evaluations on real datasets suggest that Celloc can reconstruct cellular spatial structures with various cell types across different tissues and histological regions.