Spatially Resolved Transcriptomics in Life Sciences Research

Figure 1

A cross-section of human skeletal muscle showing muscle cells and a nerve nearby. Stained with hematoxylin and eosin.
A cross-section of skeletal muscle tissue showing muscle cells and a small nerve.

Figure 2

alt text for accessibility purposes
Signaling between adjacent cells. The Notch protein functions as a receptor for ligands that activate or inhibit such receptors. Receptor-ligand interactions ground cell signaling and communication, often requiring close proximity between cells.

Figure 3

a general schematic showing fluorescence in situ hybridization
Overview of fluorescence in situ hybridization (FISH).

Figure 4

alt text for accessibility purposes
Schematic representation of multiplexed error-robust FISH (MERFISH). Binary codes assigned to mRNA species of interest, where “1” represents a short fluorescent DNA probe. b, Consecutive hybridization rounds, bleaching in between is implied, but not shown for clarity. At the end of the sixth round, it is possible to tell different mRNAs apart due to the decoded combinations of “1” and “0”.

Figure 5

a general schematic showing in situ sequencing Adapted from Spatial Transcriptomics Overview by SlifertheRyeDragon. Image created with Biorender.com. Public domain, via Wikimedia Commons CC BY-SA 4.0 DEED

Figure 6

a general schematic showing laser-capture microdissection Adapted from Spatial Transcriptomics Overview by SlifertheRyeDragon. Image created with Biorender.com. Public domain, via Wikimedia Commons CC BY-SA 4.0 DEED

Figure 7

a general schematic showing a microfluidics workflow with DBit-seq on formalin-fixed paraffin embedded (FFPE) tissue Adapted from Liu Y, Enninful A, Deng Y, & Fan R (2020). Spatial transcriptome sequencing of FFPE tissues at cellular level. Preprint. CC BY-SA 4.0 DEED

Figure 8

A graphic showing printed spots on a glass slide that are identified by a barcode and that contain primers to capture mRNA from the tissue laid on top of them
A sequencing-based spatial transcriptomics method using printed spots on a slide.

Figure 9

a general schematic showing the Visium technology
Overview of Visium technology with fresh-frozen (FF) or formalin-fixed paraffin embedded (FFPE) tissue. Source: 10x Genomics Visium

Data and Study Design

Figure 1

A graphic showing printed spots on a glass slide that are identified by a barcode and that contain oligonucleotides to capture messenger RNA from the tissue laid on top of them
Visium spatial gene expression slide

Figure 2

A Visium spatial transcriptomics workflow with fresh-frozen tissue Graphic from Grant application resources for Visium products at 10X Genomics

Figure 3

A graphic showing printed spots on a glass slide that are identified by a barcode and that contain primers to capture messenger RNA from the tissue laid on top of them
Sequencing data is mapped back to spots on the slide and compared to an image of the tissue to localize expression

Figure 4

An example of spatial transcriptomics data showing genes in rows and barcodes (spots) in columns
Spatial transcriptomics data include genes in rows and barcodes in columns

Figure 5

A human brain showing a section of dorsolateral prefrontal cortex extracted. A block of tissue containing six cortical layers and an underlying layer of white matter is excised from the section.
Tissue blocks were excised from human dorsolateral prefrontal cortex. Tissue blocks include six cortical layers and underlying white matter (wm).

Figure 6

Three Visium slides showing four spatial capture areas each. Each slide contains directly adjacent serial tissue sections for one subject. The second pair of samples contains tissue sections that are 300 microns posterior to the first pair of samples. Adapted from Maynard et al, Nat Neurosci 24, 425–436 (2021). Created with BioRender.com.

Figure 7

You plan to place samples of treated tissue on one slide and samples of the controls on another slide. What will happen when it is time for data analysis? What could you have done differently? An experiment with treated samples on one slide and control samples on another.

Figure 8

An experiment with three timepoints at 5, 10 and 15 weeks. At the end of the first 5 weeks, those samples are run through Visium. This is repeated at 10 and 15 weeks.
Three time points in an experiment

Figure 9

Four different wheel running treatments applied to 5 mice each. Treatment 1 is applied on day 1, treatment 2 on day 2, and so on.
Four different wheel running treatments each applied once per day to five mice, for a total of 20 mice treated.

Figure 10

Figure 11

A normal curve with a mean of zero showing the type 1 error rate in the far right tail and specificity in the left of the curve.
The null hypothesis states that there is no difference between treatment groups.

Figure 12

A normal curve with a mean of approximately 3 showing the type 2 error rate in the left of the curve and sensitivity (also known as statistical power) in the far right tail of the curve. The effect size is shown as the difference in means between the null and alternative hypotheses.
The alternative hypothesis states that there is a difference between treatment groups.

Data Preprocessing

Figure 1

H & E slide of sample 151673
H & E slide of sample 151673

Figure 2

Diagram showing tissue and registration spots in corners
Slide Diagram

Figure 3

Histology slide with tissue and background spots labelled
Spots identified in Tissue and Background

Figure 4

Boxplot showing lower trancript counts in background area of slide
UMI Counts in Tissue and Background

Figure 5

Boxplot showing lower numbers of genes in background area of slide
Number of Genes in Tissue and Background

Figure 6

Figure 7

Figure showing UMI counts in each spot with varying intensity across the tissue
UMI Counts in each Spot

Figure 8

Figure showing number of genes detected in each spot with varying intensity across the tissue
Number of Genes in each Spot

Remove Low-quality Spots

Figure 1

mito

Figure 2

mito

Figure 3

mito

Figure 4

Figure 5

Figure 6

Figure 7

Normalization in Spatial Transcriptomics

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

A diagram of the sttructure of a Seurat object showing the assays (Spatial & SCT) and Layers (counts, data & scale.data.
Seurat object structure

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Feature Selection, Dimensionality Reduction, and Spot Clustering

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Tissue Clustering showing the tissue layers at different resolutions and number of PCs
Tissue Clustering with Different Numbers of PCs and Clustering Resolutions

Figure 6

UMAP Clustering with Different Numbers of PCs and Clustering Resolutions
UMAP Clustering with Different Numbers of PCs and Clustering Resolutions

Figure 7

Deconvolution in Spatial Transcriptomics

Figure 1

alt text for accessibility purposesZhang et al, Comput Struct Biotechnol J 21, 176–184 (2023) CC BY-NC-ND 4.0

Figure 2

Figure 3

Figure 4

Differential Expression Testing

Figure 1

Figure 2

Figure 3

Figure 4

Moran's I statistic quantifies spatial correlation.
Moran’s I statistic quantifies spatial correlation. Top Left: Checkerboard pattern results in negative Moran’s I, indicating anti-correlation. Top Right: Linear gradient shows a high positive Moran’s I, reflecting a strong spatial gradient. Bottom Left: Random pattern leads to a Moran’s I near zero, suggesting no significant spatial autocorrelation. Bottom Right: ‘Ink blot’ pattern demonstrates positive autocorrelation, indicative of a clustered or spreading pattern. Relationships are calculated using direct, equally weighted neighbors, normalized for each cell.

Figure 5

Figure 6

Figure 7

Genes with the lowest adjusted p-values from differential expression analysis with DESeq2
Visium spatial gene expression slide

Figure 8

Spatial distribution of the most DE genes for all samples

Putting it all Together

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5