Contents

1 Introduction

tidySpatialExperiment provides a bridge between the SpatialExperiment package and the tidyverse ecosystem. It creates an invisible layer that allows you to interact with a SpatialExperiment object as if it were a tibble; enabling the use of functions from dplyr, tidyr, ggplot2 and plotly. But, underneath, your data remains a SpatialExperiment object.

tidySpatialExperiment also provides five additional utility functions.

1.1 Resources

If you would like to learn more about tidySpatialExperiment and tidyomics, the following links are a good place to start:

The tidyomics ecosystem also includes packages for:

  • Working with genomic features:

    • plyranges, for tidy manipulation of genomic range data.
    • nullranges, for tidy generation of genomic ranges representing the null hypothesis.
    • plyinteractions, for tidy manipulation of genomic interaction data.
  • Working with transcriptomic features:

  • Working with cytometry features:

    • tidytof, for tidy manipulation of high-dimensional cytometry data.
  • And a few associated packages:

    • tidygate, for manual gating of points in space.
    • tidyheatmap, for modular heatmap contruction.

1.2 Functions and utilities

Package Functions available
SpatialExperiment All
dplyr arrange,bind_rows, bind_cols, distinct, filter, group_by, summarise, select, mutate, rename, left_join, right_join, inner_join, slice, sample_n, sample_frac, count, add_count
tidyr nest, unnest, unite, separate, extract, pivot_longer
ggplot2 ggplot
plotly plot_ly
Utility Description
as_tibble Convert cell data to a tbl_df
join_features Append feature data to cell data
aggregate_cells Aggregate cell-feature abundance into a pseudobulk SummarizedExperiment object
rectangle Select cells in a rectangular region of space
ellipse Select cells in an elliptical region of space
gate_spatial
gate_programmatic

1.3 Installation

You can install the stable version of tidySpatialExperiment from Bioconductor.

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("tidySpatialExperiment")

Or, you can install the development version of tidySpatialExperiment from GitHub.

if (!requireNamespace("pak", quietly = TRUE))
    install.packages("pak")

pak::pak("william-hutchison/tidySpatialExperiment")

1.4 Load data

Here, we attach tidySpatialExperiment and an example SpatialExperiment object.

# Load example SpatialExperiment object
library(tidySpatialExperiment)
example(read10xVisium)

1.5 SpatialExperiment-tibble abstraction

A SpatialExperiment object represents assay-feature values as rows and cells as columns. Additional information about the cells is stored in the reducedDims, colData and spatialCoords slots.

tidySpatialExperiment provides a SpatialExperiment-tibble abstraction, representing cells as rows and cell data as columns, in accordance with the tidy observation-variable convention. The cell data is made up of information stored in the colData and spatialCoords slots.

The default view is now of the SpatialExperiment-tibble abstraction.

spe
## # A SpatialExperiment-tibble abstraction: 50 × 7
## # Features = 50 | Cells = 50 | Assays = counts
##    .cell              in_tissue array_row array_col sample_id pxl_col_in_fullres
##    <chr>              <lgl>         <int>     <int> <chr>                  <int>
##  1 AAACAACGAATAGTTC-1 FALSE             0        16 section1                2312
##  2 AAACAAGTATCTCCCA-1 TRUE             50       102 section1                8230
##  3 AAACAATCTACTAGCA-1 TRUE              3        43 section1                4170
##  4 AAACACCAATAACTGC-1 TRUE             59        19 section1                2519
##  5 AAACAGAGCGACTCCT-1 TRUE             14        94 section1                7679
##  6 AAACAGCTTTCAGAAG-1 FALSE            43         9 section1                1831
##  7 AAACAGGGTCTATATT-1 FALSE            47        13 section1                2106
##  8 AAACAGTGTTCCTGGG-1 FALSE            73        43 section1                4170
##  9 AAACATGGTGAGAGGA-1 FALSE            62         0 section1                1212
## 10 AAACATTTCCCGGATT-1 FALSE            61        97 section1                7886
## # ℹ 40 more rows
## # ℹ 1 more variable: pxl_row_in_fullres <int>

But, our data maintains its status as a SpatialExperiment object. Therefore, we have access to all SpatialExperiment functions.

spe |>
    colData() |>
    head()
## DataFrame with 6 rows and 4 columns
##                    in_tissue array_row array_col   sample_id
##                    <logical> <integer> <integer> <character>
## AAACAACGAATAGTTC-1     FALSE         0        16    section1
## AAACAAGTATCTCCCA-1      TRUE        50       102    section1
## AAACAATCTACTAGCA-1      TRUE         3        43    section1
## AAACACCAATAACTGC-1      TRUE        59        19    section1
## AAACAGAGCGACTCCT-1      TRUE        14        94    section1
## AAACAGCTTTCAGAAG-1     FALSE        43         9    section1
spe |> 
    spatialCoords() |>
    head()
##                    pxl_col_in_fullres pxl_row_in_fullres
## AAACAACGAATAGTTC-1               2312               1252
## AAACAAGTATCTCCCA-1               8230               7237
## AAACAATCTACTAGCA-1               4170               1611
## AAACACCAATAACTGC-1               2519               8315
## AAACAGAGCGACTCCT-1               7679               2927
## AAACAGCTTTCAGAAG-1               1831               6400
spe |>
    imgData()
## DataFrame with 1 row and 4 columns
##     sample_id    image_id   data scaleFactor
##   <character> <character> <list>   <numeric>
## 1    section1      lowres   ####   0.0510334

2 Integration with the tidyverse ecosystem

2.1 Manipulate with dplyr

Most functions from dplyr are available for use with the SpatialExperiment-tibble abstraction. For example, filter() can be used to filter cells by a variable of interest.

spe |>
   filter(array_col < 5)
## # A SpatialExperiment-tibble abstraction: 3 × 7
## # Features = 50 | Cells = 3 | Assays = counts
##   .cell              in_tissue array_row array_col sample_id pxl_col_in_fullres
##   <chr>              <lgl>         <int>     <int> <chr>                  <int>
## 1 AAACATGGTGAGAGGA-1 FALSE            62         0 section1                1212
## 2 AAACGAAGATGGAGTA-1 FALSE            58         4 section1                1487
## 3 AAAGAATGACCTTAGA-1 FALSE            64         2 section1                1349
## # ℹ 1 more variable: pxl_row_in_fullres <int>

And mutate can be used to add new variables, or modify the value of an existing variable.

spe |>
    mutate(in_region = c(in_tissue & array_row < 10))
## # A SpatialExperiment-tibble abstraction: 50 × 8
## # Features = 50 | Cells = 50 | Assays = counts
##    .cell    in_tissue array_row array_col sample_id in_region pxl_col_in_fullres
##    <chr>    <lgl>         <int>     <int> <chr>     <lgl>                  <int>
##  1 AAACAAC… FALSE             0        16 section1  FALSE                   2312
##  2 AAACAAG… TRUE             50       102 section1  FALSE                   8230
##  3 AAACAAT… TRUE              3        43 section1  TRUE                    4170
##  4 AAACACC… TRUE             59        19 section1  FALSE                   2519
##  5 AAACAGA… TRUE             14        94 section1  FALSE                   7679
##  6 AAACAGC… FALSE            43         9 section1  FALSE                   1831
##  7 AAACAGG… FALSE            47        13 section1  FALSE                   2106
##  8 AAACAGT… FALSE            73        43 section1  FALSE                   4170
##  9 AAACATG… FALSE            62         0 section1  FALSE                   1212
## 10 AAACATT… FALSE            61        97 section1  FALSE                   7886
## # ℹ 40 more rows
## # ℹ 1 more variable: pxl_row_in_fullres <int>

2.2 Tidy with tidyr

Most functions from tidyr are also available. Here, nest() is used to group the data by sample_id, and unnest() is used to ungroup the data.

# Nest the SpatialExperiment object by sample_id
spe_nested <-
    spe |> 
    nest(data = -sample_id)

# View the nested SpatialExperiment object
spe_nested
## # A tibble: 1 × 2
##   sample_id data           
##   <chr>     <list>         
## 1 section1  <SptlExpr[,50]>
# Unnest the nested SpatialExperiment objects
spe_nested |>
    unnest(data)
## # A SpatialExperiment-tibble abstraction: 50 × 7
## # Features = 50 | Cells = 50 | Assays = counts
##    .cell              in_tissue array_row array_col sample_id pxl_col_in_fullres
##    <chr>              <lgl>         <int>     <int> <chr>                  <int>
##  1 AAACAACGAATAGTTC-1 FALSE             0        16 section1                2312
##  2 AAACAAGTATCTCCCA-1 TRUE             50       102 section1                8230
##  3 AAACAATCTACTAGCA-1 TRUE              3        43 section1                4170
##  4 AAACACCAATAACTGC-1 TRUE             59        19 section1                2519
##  5 AAACAGAGCGACTCCT-1 TRUE             14        94 section1                7679
##  6 AAACAGCTTTCAGAAG-1 FALSE            43         9 section1                1831
##  7 AAACAGGGTCTATATT-1 FALSE            47        13 section1                2106
##  8 AAACAGTGTTCCTGGG-1 FALSE            73        43 section1                4170
##  9 AAACATGGTGAGAGGA-1 FALSE            62         0 section1                1212
## 10 AAACATTTCCCGGATT-1 FALSE            61        97 section1                7886
## # ℹ 40 more rows
## # ℹ 1 more variable: pxl_row_in_fullres <int>

2.3 Plot with ggplot2

The ggplot() function can be used to create a plot directly from a SpatialExperiment object. This example also demonstrates how tidy operations can be combined to build up more complex analysis.

spe |>
    filter(sample_id == "section1" & in_tissue) |>
    
    # Add a column with the sum of feature counts per cell
    mutate(count_sum = purrr::map_int(.cell, ~
        spe[, .x] |> 
            counts() |> 
            sum()
      )) |>
    
    # Plot with tidySpatialExperiment and ggplot2
    ggplot(aes(x = reorder(.cell, count_sum), y = count_sum)) +
    geom_point() +
    coord_flip()

2.4 Plot with plotly

The plot_ly() function can also be used to create a plot from a SpatialExperiment object.

spe |>
    filter(sample_id == "section1") |>
    plot_ly(
        x = ~ array_col, 
        y = ~ array_row, 
        color = ~ in_tissue, 
        type = "scatter"
    )

3 Utilities

3.1 Append feature data to cell data

The tidyomics ecosystem places an emphasis on interacting with cell data. To interact with feature data, the join_features() function can be used to append assay-feature values to cell data.

# Join feature data in wide format, preserving the SpatialExperiment object
spe |>
    join_features(features = c("ENSMUSG00000025915", "ENSMUSG00000042501"), shape = "wide") |> 
    head()
## # A SpatialExperiment-tibble abstraction: 50 × 9
## # Features = 6 | Cells = 50 | Assays = counts
##    .cell              in_tissue array_row array_col sample_id ENSMUSG00000025915
##    <chr>              <lgl>         <int>     <int> <chr>                  <dbl>
##  1 AAACAACGAATAGTTC-1 FALSE             0        16 section1                   0
##  2 AAACAAGTATCTCCCA-1 TRUE             50       102 section1                   0
##  3 AAACAATCTACTAGCA-1 TRUE              3        43 section1                   0
##  4 AAACACCAATAACTGC-1 TRUE             59        19 section1                   0
##  5 AAACAGAGCGACTCCT-1 TRUE             14        94 section1                   0
##  6 AAACAGCTTTCAGAAG-1 FALSE            43         9 section1                   0
##  7 AAACAGGGTCTATATT-1 FALSE            47        13 section1                   0
##  8 AAACAGTGTTCCTGGG-1 FALSE            73        43 section1                   0
##  9 AAACATGGTGAGAGGA-1 FALSE            62         0 section1                   0
## 10 AAACATTTCCCGGATT-1 FALSE            61        97 section1                   0
## # ℹ 40 more rows
## # ℹ 3 more variables: ENSMUSG00000042501 <dbl>, pxl_col_in_fullres <int>,
## #   pxl_row_in_fullres <int>
# Join feature data in long format, discarding the SpatialExperiment object
spe |>
    join_features(features = c("ENSMUSG00000025915", "ENSMUSG00000042501"), shape = "long") |> 
    head()
## tidySpatialExperiment says: A data frame is returned for independent data 
##               analysis.
## # A tibble: 6 × 7
##   .cell       in_tissue array_row array_col sample_id .feature .abundance_counts
##   <chr>       <lgl>         <int>     <int> <chr>     <chr>                <dbl>
## 1 AAACAACGAA… FALSE             0        16 section1  ENSMUSG…                 0
## 2 AAACAACGAA… FALSE             0        16 section1  ENSMUSG…                 0
## 3 AAACAAGTAT… TRUE             50       102 section1  ENSMUSG…                 0
## 4 AAACAAGTAT… TRUE             50       102 section1  ENSMUSG…                 1
## 5 AAACAATCTA… TRUE              3        43 section1  ENSMUSG…                 0
## 6 AAACAATCTA… TRUE              3        43 section1  ENSMUSG…                 0

3.2 Aggregate cells

Sometimes, it is necessary to aggregate the gene-transcript abundance from a group of cells into a single value. For example, when comparing groups of cells across different samples with fixed-effect models.

The aggregate_cells() function can be used to aggregate cells by a specified variable and assay, returning a SummarizedExperiment object.

spe |>
    aggregate_cells(in_tissue, assays = "counts")
## class: SummarizedExperiment 
## dim: 50 2 
## metadata(0):
## assays(1): counts
## rownames(50): ENSMUSG00000002459 ENSMUSG00000005886 ...
##   ENSMUSG00000104217 ENSMUSG00000104328
## rowData names(1): feature
## colnames(2): FALSE TRUE
## colData names(3): in_tissue .aggregated_cells sample_id

3.3 Elliptical and rectangular region selection

The ellipse() and rectangle() functions can be used to select cells by their position in space.

spe |>
    filter(sample_id == "section1") |>
    mutate(in_ellipse = ellipse(array_col, array_row, c(20, 40), c(20, 20))) |>
    ggplot(aes(x = array_col, y = array_row, colour = in_ellipse)) +
    geom_point()

3.4 Interactive gating

For the interactive selection of cells in space, tidySpatialExperiment experiment provides gate(). This function uses tidygate, shiny and plotly to launch an interactive plot overlaying cells in position with image data. Additional parameters can be used to specify point colour, shape, size and alpha, either with a column in the SpatialExperiment object or a constant value.

spe_gated <- 
  spe |>
  gate(colour = "in_tissue", alpha = 0.8)