Size matters - the impact of nucleus size on results from spatial transcriptomics

Abstract Background: Visium Spatial Gene Expression (ST) is a method combining histological spatial information with transcriptomics profiles directly from tissue sections. The use of spatial information has made it possible to discover new modes of gene expression regulations. However, in the ST experiment, the nucleus size of cells may exceed the thickness of a tissue slice. This may, in turn, negatively affect comprehensive capturing the transcriptomics profile in a single slice, especially for tissues having large differences in the size of nuclei. Methods: We applied Consecutive Slices Data Integration (CSDI) to unveil accurate spot classification and clustering, followed by the deconvolution of spatial transcriptomic spots in human postmortem brains. We integrated spatial transcriptomic profiles with single nuclei RNA-seq data. We used histological information as a reference, to asses cell identification. Results: We observed significant improvement in cell recognition and spot classification. Apart from the escalated number of defined clusters representing neuronal layers, the pattern of clusters in consecutive sections was concordant only after CSDI. Additionally, the assigned cell labels to spots match the histological pattern of tissue sections after CSDI. Conclusion: CSDI can be applied to investigate consecutive sections studied with ST in the human cerebral cortex, avoiding misinterpretation of spot clustering and annotation, increasing accuracy of cell recognition as well as improvement in uncovering the layers of grey matter in the human brain.