In ellispatrick/spicyR: Spatial analysis of in situ cytometry data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  warning = FALSE
)
library(BiocStyle)

Installation

if (!require("BiocManager")) {
  install.packages("BiocManager")
}
BiocManager::install("spicyR")

# load required packages
library(spicyR)
library(ggplot2)
library(SingleCellExperiment)
library(SpatialExperiment)
library(imcRtools)

Overview

This guide will provide a step-by-step guide on how mixed effects models can be applied to multiple segmented and labelled images to identify how the localisation of different cell types can change across different conditions. Here, the subject is modelled as a random effect, and the different conditions are modelled as a fixed effect.

Example data

Here, we use a subset of the Damond et al., 2019 imaging mass cytometry dataset. We will compare the spatial distributions of cells in the pancreatic islets of individuals with early onset diabetes and healthy controls.

diabetesData_SCE is a SingleCellExperiment object containing single-cell data of 160 images from 8 subjects, with 20 images per subject.

data("diabetesData_SCE")
diabetesData_SCE

In this data set, cell types include immune cell types (B cells, naive T cells, T Helper cells, T cytotoxic cells, neutrophils, macrophages) and pancreatic islet cells (alpha, beta, gamma, delta).

Mixed Effects Modelling

To investigate changes in localisation between two different cell types, we measure the level of localisation between two cell types by modelling with the L-function. Specifically, the mean difference between the obtained function and the theoretical function is used as a measure for the level of localisation. Differences of this statistic between two conditions is modelled using a weighted mixed effects model, with condition as the fixed effect and subject as the random effect.

Testing for change in localisation for a specific pair of cells

Firstly, we can test whether one cell type tends to be more localised with another cell type in one condition compared to the other. This can be done using the spicy() function, where we include condition, and subject.

In this example, we want to see whether or not Delta cells (to) tend to be found around Beta cells (from) in onset diabetes images compared to non-diabetic images. However, given that there are 3 conditions, we can specify the desired conditions by setting the order of our condition factor. spicy() will choose the first level of the factor as the base condition and the second level as the comparison condition. spicy() will also naturally coerce the condition column into a factor if not already a factor.

spicyTestPair <- spicy(
  diabetesData_SCE,
  condition = "stage",
  subject = "case",
  from = "beta",
  to = "delta"
)

topPairs(spicyTestPair)

We obtain a spicy object which details the results of the mixed effects modelling performed. As the coefficient in spicyTest is positive, we find that delta cells cells are significantly less likely to be found near beta cells in the onset diabetes group compared to the non-diabetic control.

Test for change in localisation for all pairwise cell combinations

Here, we can perform what we did above for all pairwise combinations of cell types by excluding the from and to parameters from spicy().

spicyTest <- spicy(
  diabetesData_SCE,
  condition = "stage",
  subject = "case"
)

topPairs(spicyTest)

We can also examine the L-function metrics of individual images by using the convenient bind() function on our spicyTest results object.

bind(spicyTest)[1:5, 1:5]

Again, we obtain a spicy object which outlines the result of the mixed effects models performed for each pairwise combination of cell types.

We can represent this as a bubble plot using the signifPlot() function by providing it the spicy object obtained. Here, we can observe that the most significant relationships occur between pancreatic beta and delta cells, suggesting that the 2 cell types are far more localised during diabetes onset compared to non-diabetics.

signifPlot(
  spicyTest,
  breaks = c(-3, 3, 1),
  marksToPlot = c(
    "alpha", "beta", "gamma", "delta",
    "B", "naiveTc", "Th", "Tc", "neutrophil", "macrophage"
  )
)

If we're interested and wish to examine a specific cell type-cell type relationship in more detail, we can use spicyBoxPlot, specifying the relationship we want to examine.

In the bubble plot above, we can see that the most significant relationship is between beta and delta islet cells in the pancreas. To examine this further, we can specify either from = beta and to = delta or rank = 1 parameters in spicyBoxPlot.

spicyBoxPlot(results = spicyTest,
             # from = "beta",
             # to = "delta",
             rank = 1)

Mixed effects modelling for custom metrics

spicyR can also be applied to custom distance or abundance metrics. Here, we provide an example where we apply the spicy function to a custom abundance metric.

We first obtain the custom abundance metric by converting the a SpatialExperiment object from the existing SingleCellExperiment object. A KNN interactions graph is then generated with the function buildSpatialGraph from the imcRtools package. This generates a colPairs object inside of the SpatialExperiment object. spicyR provides the function convPairs for converting a colPairs object stored within a SingleCellExperiment object into an abundance matrix by effectively calculating the average number of nearby cells types for every cell type. For example, if there exists on average 5 neutrophils for every macrophage in image 1, the column neutrophil__macrophage would have a value of 5 for image 1.

spicy can take any input of pairwise cell type combinations across multiple images and run a mixed effects model to determine collective differences across conditions.

diabetesData_SPE <- SpatialExperiment(diabetesData_SCE,
                                      colData = colData(diabetesData_SCE)) 

spatialCoords(diabetesData_SPE) <- data.frame(colData(diabetesData_SPE)$x, colData(diabetesData_SPE)$y) |> as.matrix()
spatialCoordsNames(diabetesData_SPE) <- c("x", "y")

diabetesData_SPE <- imcRtools::buildSpatialGraph(diabetesData_SPE, img_id = "imageID", type = "knn", k = 20, coords = c("x", "y"))

pairAbundances <- convPairs(diabetesData_SPE,
                  colPair = "knn_interaction_graph")

head(pairAbundances["delta__delta"])

spicy can take any input of pairwise cell type combinations across multiple images and run a mixed effects model to determine collective differences across conditions. To check out other custom distance metrics which can be used, feel free to check out the Statial package.

spicyTestColPairs <- spicy(
  diabetesData_SPE,
  condition = "stage",
  subject = "case",
  alternateResult = pairAbundances,
  weights = FALSE
)

topPairs(spicyTestColPairs)

Again, we can present this spicy object as a bubble plot using the signifPlot() function by providing it with the spicy object.

signifPlot(
  spicyTestColPairs,
  marksToPlot = c(
    "alpha", "acinar", "ductal", "naiveTc", "neutrophil", "Tc",
    "Th", "otherimmune"
  )
)

sessionInfo()

sessionInfo()

ellispatrick/spicyR documentation built on April 27, 2024, 10:14 a.m.