Deconvolution Tactics and Normalization in Renal Spatial Transcriptomics.

Ricardo Melo Ferreira,Michael T Eadon,Benjamin J Freije

doi:10.3389/fphys.2021.812947

Ricardo Melo Ferreira, Michael T Eadon + Show 1 more

Open Access

https://doi.org/10.3389/fphys.2021.812947

Copy DOI

Abstract

The kidney is composed of heterogeneous groups of epithelial, endothelial, immune, and stromal cells, all in close anatomic proximity. Spatial transcriptomic technologies allow the interrogation of in situ expression signatures in health and disease, overlaid upon a histologic image. However, some spatial gene expression platforms have not yet reached single-cell resolution. As such, deconvolution of spatial transcriptomic spots is important to understand the proportion of cell signature arising from these varied cell types in each spot. This article reviews the various deconvolution strategies discussed in the 2021 Indiana O’Brien Center for Microscopy workshop. The unique features of Seurat transfer score methodology, SPOTlight, Robust Cell Type Decomposition, and BayesSpace are reviewed. The application of normalization and batch effect correction across spatial transcriptomic samples is also discussed.

Highlights

Spatial transcriptomics was selected as Nature’s Method of the year in 2020 (Marx, 2021)
As presented at the 2021 O’Brien Center for Microscopy workshop, Spatial Transcriptomics (ST) represents a powerful tool to reveal in situ transcript expression associated with histopathologic features
This indicates that technical variation in our samples can be modeled by sequencing depth alone. These results suggest SCTransform may be a useful tool for removing intersample technical variation in ST datasets. This brief review presents the result of four common deconvolution techniques and a common normalization procedure applied to the human kidney, as discussed in the 2021 O’Brien Center for Microscopy workshop

Summary

Introduction

Spatial transcriptomics was selected as Nature’s Method of the year in 2020 (Marx, 2021). Visium Spatial Gene Expression (VSGE) platform has a spot size of 55 μm and resolution of 100 μm, which invariably encompasses multiple cells within a single spot. These classes of cell types align very well with the underlying histology of the human kidney (Melo Ferreira et al, 2021).

Results

Conclusion