Impact of single-cell RNA reference selection for the deconvolution of breast cancer spatial transcriptomics datasets

BackgroundSpot-based Spatial transcriptomics (ST) allows for unbiased gene expression analysis within tissue architecture, overcoming the limitations of single-cell RNA sequencing (scRNA-seq) by preserving spatial context. However, the high spatial resolution in ST leads to cellular heterogeneity within spots, requiring computational deconvolution to infer cellular compositions. While scRNA-seq serves as a key reference for deconvolution, the impact of reference composition on its accuracy is still unclear. In this study, we systematically evaluate the impact of reference selection for cellular deconvolution and give helpful guide-lines to researchers to address this task. MethodsWe systematically evaluate the impact of single-cell reference selection for global and cell type-specific deconvolution performances in primary and metastatic breast cancer ST datasets. Focusing on state-of-the-art deconvolution tools Cell2location and Robust Cell Type Decomposition (RCTD), we assess the influence of varying reference sizes, cell-type distributions, reference-ST pairings, as well as the usage of large breast cancer and cross-cancer atlases. ResultsOur findings demonstrate that even small reference datasets can yield accurate deconvolution results, with RCTD and Cell2location exhibiting similar performances across cell-types. Heterogeneous entities, such as myeloids, presented greater challenges for accurate deconvolution compared to more distinct ones, such as epithelial cells. Spatial domains associated with prominent cell types like cancer and stromal cells were detected, although their contributions were systematically under- or over-estimated Additionally, we found that perfect reference-ST matching enhances deconvolution accuracy compared to the usage of cross-patient references. Finally, large breast cancer single-cell atlases com-posed of a diverse set of patients were also able to provide reliable deconvolution results. ConclusionThis study provides important insights into optimizing reference selection for ST deconvolution, highlighting the strengths and limitations of current computational tools in addressing cellular heterogeneity within spatial transcriptomics datasets.