CellContrast: Reconstructing spatial relationships in single-cell RNA sequencing data via deep contrastive learning

Spatial transcriptomics (ST) could revolutionize our understanding of biological tissues and processes. However, its usage could be associated with high costs and technical challenges. Meanwhile, single-cell RNA sequencing (SC) technologies have matured and there is currently a large amount of data available. These rich SC datasets hold great potential for biological discoveries, but they lack spatial relationship information between cells. CellContrast is a computational method that uses gene expression data and a model trained with ST data to reconstruct the spatial relationships between cells. CellContrast revealed that spatial information can still be transferred to the SC data even if the SC and ST information do not align perfectly. Mining of existing SC datasets coupled with limited ST information can maximize the value of these two resources and contribute to further understanding of cell systems.