escher

name: bioconductor-escherdescription: The creation of effective visualizations is a fundamental component of data analysis. In biomedical research, new challenges are emerging to visualize multi-dimensional data in a 2D space, but current data visualization tools have limited capabilities. To address this problem, we leverage Gestalt principles to improve the design and interpretability of multi-dimensional data in 2D data visualizations, layering aesthetics to display multiple variables. The proposed visualization can be applied towhen_to_use: Use when: The creation of effective visualizations is a fundamental component of data analysis; install bioc using install; install github using install_github, install; data.import using Visium_humanDLPFC; create plot using make_escheR; creat fill using add_fill. Not for: Requires R ≥ 4.3user-invocable: false

escheR

Workflows

Standard Workflow

Visualize multi-dimensional spatial transcriptomics data from a SpatialExperiment or SingleCellExperiment object by adding layered fill, ground, and symbol aesthetics.

library(escheR)
library(STexampleData)

# 1. Read the input SpatialExperiment object
spe <- Visium_humanDLPFC()
spe <- spe[, spe$in_tissue == 1]
spe <- spe[, !is.na(spe$ground_truth)]

# 2. Create plot and add layered aesthetics
p <- make_escheR(spe) |>
  add_fill(var = "cell_count") |>
  add_ground(var = "ground_truth") |>
  add_symbol(var = "ground_truth", size = 0.2)

# 3. Save the generated plots to the working directory
ggplot2::ggsave(filename = "standard_plot.pdf", plot = p)

Input: A SpatialExperiment object spe. Output: A saved PDF plot showing layered spatial aesthetics.

Binned Spatial Visualization

Visualize high-density spatial transcriptomics data using point binning to avoid overplotting.

library(escheR)
library(STexampleData)

# 1. Read the user input file
spe <- Visium_humanDLPFC()
spe <- spe[, spe$in_tissue == 1]
spe <- spe[, !is.na(spe$ground_truth)]

# 2. Ensure the SpatialExperiment object has the required columns and factor types
spe$counts_MOBP <- counts(spe)[which(rowData(spe)$gene_name == "MOBP"), ]
spe$ground_truth <- factor(spe$ground_truth)

# 3. Create binned plot and save to a PDF file
p <- make_escheR(spe, dimred = "PCA") |>
  add_ground_bin(var = "ground_truth") |>
  add_fill_bin(var = "counts_MOBP") +
  scale_fill_gradient(low = "white", high = "black", name = "MOBP Count") +
  scale_color_discrete(name = "Spatial Domains")

ggplot2::ggsave(filename = "binned_plot.pdf", plot = p)

Input: A SpatialExperiment object spe with PCA reduced dimensions. Output: A saved PDF plot with binned spatial points.

Dataframe Visualization

Visualize multi-dimensional data from a generic R data.frame (e.g., exported from Seurat) using custom x and y coordinates.

library(escheR)

# Create a generic data.frame with coordinates and metadata
df <- data.frame(
  x = c(1, 2, 3),
  y = c(4, 5, 6),
  groups = c("A", "B", "A")
)

# Call generic function for make_escheR.data.frame
p <- make_escheR(object = df, .x = df$x, .y = df$y) |>
  add_fill(var = "groups")

Input: A standard R data.frame with coordinate columns. Output: A ggplot object.

When to Use

• Visualizing multi-dimensional spatial transcriptomics data (e.g., from 10x Visium, seqFISH, or Slide-seq V2) by layering aesthetics.
• Plotting dimensionality reduced embeddings (e.g., PCA) from a SpatialExperiment or SingleCellExperiment using make_escheR(spe, dimred = "PCA").
• Visualizing high-density spatial data using point binning with add_ground_bin() and add_fill_bin().
• Visualizing spatial or embedding data from a generic R data.frame using make_escheR(object, .x, .y).

When NOT to Use

• For complex multi-sample joint analysis or alignment, use packages like Seurat or Giotto because escheR operates on single samples by design.
• For advanced cell-cell communication or spatial ligand-receptor analysis, use specialized packages as escheR is strictly a visualization tool.

Data Requirements

• Input can be a SpatialExperiment, SingleCellExperiment, or a base R data.frame.
• For SpatialExperiment, spatial coordinates must be present (e.g., accessed via spatialCoords(spe)).
• For data.frame input, explicit numeric vectors for x and y coordinates must be provided to .x and .y.

Key Parameters

• object: The input data object (SpatialExperiment, SingleCellExperiment, or data.frame).
• dimred (NULL): Name of the reduced dimension slot to use for coordinates instead of spatial coordinates.
• .x: Numeric vector of x coordinates (for data.frame input).
• .y: Numeric vector of y coordinates (for data.frame input).
• var: Character string specifying the column name in colData or data.frame to map to an aesthetic layer.
• stroke (0.5): Border stroke size for the ground layer.
• size (0.2): Size of the symbols in the symbol layer.

Best Practices

• Always apply add_fill() as the first layer before other add_* functions to achieve the best visual effect due to the layering mechanism.
• When using binning to avoid overplotting, apply add_fill_bin() after add_ground_bin() for a better visualization outcome.
• Trim down colData(spe) before piping into make_escheR() to reduce computation time, especially when it contains a large number of irrelevant columns.
• Use minimally overlapping color palettes (e.g., a white-to-black gradient for continuous gene expression and qualitative colors for spatial domains) to avoid visual confusion.

Common Pitfalls

• Overlapping color spaces: Using default viridis for both fill and ground makes them indistinguishable; fix by using a white-to-black gradient for continuous variables via scale_fill_gradient(low = "white", high = "black").
• Large number of categories: Visualizing too many categorical levels creates clutter; fix by subsetting levels (setting unwanted to NA) or using add_symbol() to annotate specific levels.
• Binning membership mixing: Point binning can lead to intermixing of cluster memberships; interpret carefully as the majority membership is selected as the bin label.

Alternatives

• Seurat for comprehensive single-cell and spatial transcriptomics workflows.
• scater for single-cell visualization and quality control.
• scran for single-cell data analysis and variance modeling.

Citations

• Boyi Guo and Stephanie C. Hicks (2024). escheR: unified multi-dimensional spatial visualizations. Bioinformatics Advances.

References

• Homepage: bioconductor.org/packages/escheR
• Vignette: bioconductor.org/packages/release/bioc/vignettes/escheR/inst/doc/escheR.html