Spatial Mapping of Single Cells via Correlation and Importance Between Cells and Spots

Mapping cells to spatial locations is crucial for understanding biological processes, disease mechanisms, and therapeutic strategies. However, dropout events in both spatial transcriptomics and scRNA-seq data, along with intercellular interactions and spatial dependencies among spots, challenge the accuracy of spatial mapping. This study proposes a novel method termed Spatial Mapping of single cells via Correlation and Importance between cells and spots (SM-CI). We introduce dropout handling strategies specifically designed for both spatial transcriptomics and scRNA-seq data. Imputation index sets and dropout imputation functions tailored to each data type are developed: the first effectively utilizes spatial location and gene expression information, while the second leverages data from neighboring cell types. Furthermore, we establish criteria for assessing the importance of spots (or cells) in relation to others and construct a linear programming model that integrates these criteria with correlation measures to enhance spatial mapping accuracy. Benchmarking on simulated datasets shows that SM-CI consistently outperforms existing methods across four metrics, while applications to real datasets demonstrate its effectiveness in reconstructing spatial distributions of diverse cell types across tissues. Additionally, ablation experiments validate the effectiveness of the dropout handling strategies and importance assessment criteria.