Comparative Methods for Demystifying Spatial Transcriptomics

Spatial Transcriptomics (ST), coined as the term for parallel RNA-Seq on cell populations ordered spatially on a histological tissue section, has recently become increasingly popular, especially in experiments where microfluidics-based single-cell sequencing fails, such as assays on neurons. ST platforms, like the 10x Visium technology investigated herein, therefore produce in a single experiment simultaneously thousands of RNA readouts, captured by an array of micrometer scale spots under the histological section. Therefore, a central challenge of analyzing ST experiments consists of analyzing the gene expression morphology of all spots to delineate clusters of similar cell mixtures, which are then compared to each other to identify up- or down-regulated marker genes. Moreover, another level of complexity in ST experiments, compared to traditional RNA-Seq, is imposed by staining the tissue section with protein markers of cells or cell components to identify spots providing relevant information afterward. The corresponding microscopy images need to be analyzed in addition to the RNA-Seq read mappings on the reference genome and transcriptome sequences. Focusing on the software suite provided by the Visium platform manufacturer, we break down the ST analysis pipeline into its four essential steps-the image analysis, the read alignment, the gene quantification, and the spot clustering-and compare results obtained when using reads from different subsets of spots and/or when employing alternative genome or transcriptome references. Our comparative analyses demonstrate the impact of spot selection and the choice of genome/transcriptome references on the analysis results when employing the manufacturer's pipeline.