Dear Editor,

Spatial transcriptomics (ST) makes it possible to perform RNA-seq at hundreds of precisely located spots on the surface of a histological slice (Ståhl et al., 2016). Since mRNA diffusion is minimal during tissues permeabilization and mRNA capture, the transcriptome of each spot is thought to aggregate the transcriptomes of the cells it contains. The number of cells within a spot and their transcriptional output depend on their type, state, and overall local morphology. ST shares some limitations with single-cell RNA-seq, including high dropout rate. So far, ST studies have relied on preprocessing pipelines inspired by single-cell RNA-seq studies (Ståhl et al., 2016; Asp et al., 2017; Giacomello et al., 2017; Berglund et al., 2018; Lundmark et al., 2018). These include normalization of gene-wise read counts in a cell/spot by the total number of reads collected from that cell/spot. But the number of reads obtained from a spot could reflect its cellular content or technical variation in RNA capture and amplification. Thus, whether read count normalization is warranted in the context of ST remains an open question. We addressed it by quantifying the cellular content of individual spots from image analysis and by comparing it with read counts.