Package 'spatialFDA'

Convert SpatialExperiment object to ppp object

Description

Convert SpatialExperiment object to ppp object

Usage

.dfToppp(df, marks = NULL, continuous = FALSE, window = NULL)

Arguments

df	A dataframe with the x and y coordinates from the corresponding SpatialExperiment and the ColData
marks	A vector of marks to be associated with the points, has to be either named 'cell_type' if you want to compare discrete celltypes or else continous gene expression measurements are assumed as marks.
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
window	An observation window of the point pattern of class owin.

Value

A ppp object for use with spatstat functions

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
speSub <- subset(spe, , image_number == "138")
dfSub <- .speToDf(speSub)
pp <- .dfToppp(dfSub, marks = "cell_type")

---

Compute a spatial metric on a SpatialExperiment object

Description

A function that takes a SpatialExperiment object and computes a spatial statistics function as implemented in spatstat. The output is a spatstat object.

Usage

.extractMetric(
  df,
  selection,
  fun,
  marks = NULL,
  rSeq = NULL,
  by = NULL,
  continuous = FALSE,
  window = NULL,
  ...
)

Arguments

df	A dataframe with the x and y coordinates from the corresponding SpatialExperiment and the colData
selection	the mark(s) you want to compare
fun	the spatstat function to compute on the point pattern object
marks	the marks to consider e.g. cell types
rSeq	the range of r values to compute the function over
by	the spe colData variable(s) to add to the meta data
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
window	a observation window for the point pattern of class owin.
...	Other parameters passed to spatstat.explore functions

Value

a spatstat metric object with the fov number, the number of points and the centroid of the image

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
speSub <- subset(spe, , image_number == "138")
dfSub <- .speToDf(speSub)
metricRes <- .extractMetric(dfSub, c("alpha", "Tc"),
    fun = "Gcross",
    marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
    by = c("patient_stage", "patient_id", "image_number")
)

---

load Example dataset from Damond et al. (2019)

Description

load Example dataset from Damond et al. (2019)

Usage

.loadExample(full = FALSE)

Arguments

full	a boolean indicating whether to load the entire Damond et al. (2019) or only a subset

Value

A SpatialExperiment object as uploaded to ExperimentHub()

Examples

# retrieve the Damond et al. (2019) dataset
spe <- .loadExample()

---

Transform a SpatialExperiment into a dataframe

Description

Transform a SpatialExperiment into a dataframe

Usage

.speToDf(spe)

Arguments

spe	A SpatialExperiment object subset to a single image

Value

A dataframe with the x and y coordinates from the corresponding SpatialExperiment and the colData

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
speSub <- subset(spe, , image_number == "138")
dfSub <- .speToDf(speSub)

---

Calculate cross spatial metrics for all combinations per FOV

Description

A function that takes a SpatialExperiment object as input and calculates a cross spatial metric as implemented by spatstat per field of view for all combinations provided by the user.

Usage

calcCrossMetricPerFov(
  spe,
  selection,
  subsetby = NULL,
  fun,
  marks = NULL,
  rSeq = NULL,
  by = NULL,
  ncores = 1,
  continuous = FALSE,
  assay = "exprs",
  ...
)

Arguments

spe	a SpatialExperiment object
selection	the mark(s) you want to compare
subsetby	the spe colData variable to subset the data by
fun	the spatstat function to compute on the point pattern object
marks	the marks to consider e.g. cell types
rSeq	the range of r values to compute the function over
by	the spe colData variable(s) to add to the meta data
ncores	the number of cores to use for parallel processing, default = 1
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
assay	the assay which is used if continuous = TRUE
...	Other parameters passed to spatstat.explore functions

Value

a dataframe of the spatstat metric objects with the radius r, the theoretical value of a Poisson process, the different border corrections the fov number, the number of points and the centroid of the image

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcCrossMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)

---

Calculate a spatial metric on a SpatialExperiment object per field of view

Description

A function that takes a SpatialExperiment object as input and calculates a spatial metric as implemented by spatstat per field of view.

Usage

calcMetricPerFov(
  spe,
  selection,
  subsetby,
  fun,
  marks = NULL,
  rSeq = NULL,
  by = NULL,
  continuous = FALSE,
  assay = "exprs",
  ncores = 1,
  verbose = TRUE,
  ...
)

Arguments

spe	a SpatialExperiment object
selection	the mark(s) you want to compare. NOTE: This is directional. c(A,B) is not the same result as c(B,A).
subsetby	the spe colData variable to subset the data by. This variable has to be provided, even if there is only one sample.
fun	the spatstat function to compute on the point pattern object
marks	the marks to consider e.g. cell types
rSeq	the range of r values to compute the function over
by	the spe colData variable(s) to add to the meta data
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
assay	the assay which is used if continuous = TRUE
ncores	the number of cores to use for parallel processing, default = 1
verbose	logical indicating whether to print all information or not
...	Other parameters passed to spatstat.explore functions

Value

a dataframe of the spatstat metric objects with the radius r, the theoretical value of a Poisson process, the different border corrections the fov number, the number of points and the centroid of the image

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)

---

Coef for functionalGam object

Description

Coef for functionalGam object

Usage

## S3 method for class 'functionalGam'
coef(object, ...)

Arguments

object	a fitted functionalGam-object
...	see coef.pffr() for options.

Value

coefficicents of the model, see coef.pffr()

Author(s)

Martin Emons, adapted from coef.pffr() by Fabian Scheipl

---

Function for Cross Spatial Inference

Description

This function is a wrappere function around spatialInference. It calculates spatialInference results either for all cell types in marks (if selection == NULL) or for a custom subset defined in selection.

Usage

crossSpatialInference(
  spe,
  selection = NULL,
  fun,
  marks = NULL,
  rSeq = NULL,
  correction,
  sample_id,
  image_id,
  condition,
  continuous = FALSE,
  assay = "exprs",
  transformation = NULL,
  eps = 0.001,
  delta = "minNnDist",
  family = stats::gaussian(link = "log"),
  verbose = TRUE,
  ncores = 1,
  ...
)

Arguments

spe	a SpatialExperiment object
selection	the mark(s) you want to compare. NOTE: This is directional. c(A,B) is not the same result as c(B,A).
fun	the spatstat function to compute on the point pattern object
marks	the marks to consider e.g. cell types
rSeq	the range of r values to compute the function over
correction	the edge correction to be applied
sample_id	the spe colData variable to mark the sample, if not NULL this will result in a mixed model estimation
image_id	the spe colData variable to mark the image
condition	the spe colData variable to mark the condition
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
assay	the assay which is used if continuous = TRUE
transformation	the transformation to be applied as exponential e.g. 1/2 for sqrt
eps	some distributional families fail if the response is zero, therefore, zeros can be replaced with a very small value eps
delta	the delta value to remove from the beginning of the spatial statistics functions. Can be reasonable if e.g. cells are always spaced by 10 µm. If set to "minNnDist" it will take the mean of the minimum nearest neighbour distance across all images for this cell type pair.
family	the distributional family for the functional GAM
verbose	logical indicating whether to print all information or not
ncores	the number of cores to use for parallel processing, default = 1
...	Other parameters passed to spatstat.explore functions for parameters concerning the spatial function calculation and to refund::pffr for the functional additive mixed model inference

Value

a list of objects created by the function spatialInference with three objects: i) the dataframe with the spatial statistics results, ii) the designmatrix of the inference and iii) the fitted pffr object

Examples

spe <- .loadExample()
#make the condition a factor variable
colData(spe)[["patient_stage"]] <- factor(colData(spe)[["patient_stage"]])
#relevel to have non-diabetic as the reference category
colData(spe)[["patient_stage"]] <- relevel(colData(spe)[["patient_stage"]],
"Non-diabetic")

selection <- c("acinar", "ductal")
resLs <- crossSpatialInference(spe, selection, fun = "Gcross",
                      marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
                      correction = "rs", sample_id = "patient_id",
                      image_id = "image_number", condition = "patient_stage",
                      algorithm = "bam",
                      ncores = 1
                  )

---

Reshaping the Result of a Cross Spatial Inference to a Dataframe

Description

Reshaping the Result of a Cross Spatial Inference to a Dataframe

Usage

extractCrossInferenceData(
  resLs,
  QCMetric = "medianMinIntensity",
  QCThreshold = 0.1
)

Arguments

resLs	a list with four objects: i) the dataframe with the spatial statistics results transformed and filtered as used for fitting, ii) the raw spatial statistics results, iii) the designmatrix of the inference and iv) the fitted pffr object v) the residual standard error per condition defined as the residual sum of squares divided by the number of datapoints - sum of the estimated degrees of freedom for the model parameters as well as other QC metrics
QCMetric	the metric to relate the quality of the fit too.
QCThreshold	the threshold on the QC metric. Depends on the function used.

Value

a dataframe for plotting with ggplot2

Examples

spe <- .loadExample()
#make the condition a factor variable
colData(spe)[["patient_stage"]] <- factor(colData(spe)[["patient_stage"]])
#relevel to have non-diabetic as the reference category
colData(spe)[["patient_stage"]] <- relevel(colData(spe)[["patient_stage"]],
"Non-diabetic")

selection <- c("acinar", "ductal")
resLs <- crossSpatialInference(spe, selection,
                      fun = "Gcross", marks = "cell_type",
                      rSeq = seq(0, 50, length.out = 50), correction = "rs",
                      sample_id = "patient_id",
                      image_id = "image_number", condition = "patient_stage",
                      algorithm = "bam",
                      ncores = 1
                  )
df <- extractCrossInferenceData(resLs)

---

General additive model with functional response

Description

A function that takes the output of a metric calculation as done by calcMetricPerFov. The data has to be prepared into the correct format for the functional analysis by the prepData function. The output is a pffr object as implemented by refund.

Usage

functionalGam(
  data,
  x,
  designmat,
  weights,
  formula,
  family = stats::gaussian(link = "log"),
  algorithm = "bam",
  bs.yindex = list(bs = "ps", k = 5, m = c(2, 1)),
  bs.int = list(bs = "ps", k = 20, m = c(2, 1)),
  sandwich = "cluster",
  discrete = TRUE,
  ...
)

Arguments

data	a dataframe with the following columns: Y = functional response; sample_id = sample ID; image_id = image ID;
x	the x-axis values of the functional response
designmat	a design matrix as defined by model.matrix()
weights	weights as the number of points per image. These weights are normalised by the mean of the weights in the fitting process
formula	the formula for the model. The colnames of the designmatrix have to correspond to the variables in the formula.
family	the distributional family as implemented in family.mgcv. For fast computation the default is set to gaussian with a log link. other interesting options can be betar and scat for more information see family.mgcv.
algorithm	algorithm to fit the refund::pffr method. defaults to bam
bs.yindex	a list specifying the spline bases for the index. See refund::pffr for more details
bs.int	a list specfying the spline bases for the global function intercept. See refund::pffr for more details
sandwich	string indicating how and if to adjust for heterscedasticity of the residuals with a sandwich correction
discrete	option to discretise the function for faster computation. default is TRUE. See mgcv::bam for more information. When using gam, this option has to be FALSE
...	Other parameters passed to pffr

Value

a fitted pffr object which inherits from gam

Examples

# load the pancreas dataset
library("tidyr")
library("dplyr")
# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
# calculate the Gcross metric for alpha and Tc cells
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross",
    marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
    c("patient_stage", "patient_id", "image_number"), ncores = 1
)
metricRes$ID <- paste0(
  metricRes$patient_stage, "|", metricRes$patient_id,
  "|", metricRes$image_number
)
dat <- prepData(metricRes, "r", "rs", sample_id = "patient_id",
    image_id = "image_number", condition = "patient_stage")

#' # drop rows with NA
dat <- dat |> drop_na()

# create a designmatrix
condition <- dat$patient_stage
# relevel the condition - can set explicit contrasts here
condition <- relevel(condition, "Non-diabetic")
designmat <- model.matrix(~condition)
# colnames don't work with the '-' sign
colnames(designmat) <- c(
    "(Intercept)", "conditionLong_duration",
    "conditionOnset"
)
# fit the model
mdl <- functionalGam(
    data = dat, x = metricRes$r |> unique(),
    designmat = designmat, weights = dat$npoints,
    formula = formula(Y ~ conditionLong_duration +
        conditionOnset + s(patient_id, bs = "re")),
        algorithm = "bam"
)
summary(mdl)

---

Functional Principal Component Analysis

Description

A function that takes as input the output of calcMetricPerFov which has to be converted into the correct format by prepData. The output is a list with the fpca.face output from refund.

Usage

functionalPCA(data, r, ...)

Arguments

data	a data object for functional data analysis containing at least the functional response $Y$.
r	the functional domain
...	Other parameters passed to fpca.sc functions

Value

a list with components of fpca.sc

Examples

# load the pancreas dataset
library("tidyr")
library("stringr")
library("dplyr")
# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
# calculate the Gcross metric for alpha and Tc cells
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross",
    marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
    c("patient_stage", "patient_id", "image_number"), ncores = 1
)
metricRes$ID <- paste0(
  metricRes$patient_stage, "|", metricRes$patient_id,
  "|", metricRes$image_number
)
# prepare data for FDA
dat <- prepData(metricRes, "r", "rs")

# drop rows with NA
dat <- dat |> drop_na()
# create meta info of the IDs
splitData <- str_split(dat$ID, "x")
dat$condition <- factor(sapply(splitData, function(x) x[1]))
dat$patient_id <- factor(sapply(splitData, function(x) x[2]))
dat$image_id <- factor(sapply(splitData, function(x) x[3]))
# calculate fPCA
mdl <- functionalPCA(
    data = dat, r = metricRes$r |> unique()
)

---

Creates a nXn plot of the cross metrics per sample

Description

Helper function for plotCrossMetricPerFov. It applies plotMetricPerFov to all n marks defined in the variable selection. This gives an nxn plot of all marks.

Usage

plotCrossFOV(
  subFov,
  theo,
  correction,
  x,
  imageId,
  ID = NULL,
  ncol = NULL,
  nrow = NULL,
  legend.position = "none",
  ...
)

Arguments

subFov	a subset of the dataframe to the respective fov
theo	logical; if the theoretical line should be plotted
correction	the border correction to plot
x	the x-axis variable to plot
imageId	the ID of the image/fov
ID	the (optional) ID for plotting combinations
ncol	the number of columns for the facet wrap
nrow	the number of rows for the facet wrap
legend.position	the position of the legend of the plot
...	Other parameters passed to ggplot2 functions

Value

a ggplot object

---

Plotting the Result of a Cross Spatial Inference

Description

Plotting the Result of a Cross Spatial Inference

Usage

plotCrossHeatmap(
  resLs,
  adj.pvalue = "BH",
  coefficientsToPlot = NULL,
  QCThreshold = 1e-05,
  QCMetric = "medianMinIntensity",
  ...
)

Arguments

resLs	a list with four objects: i) the dataframe with the spatial statistics results transformed and filtered as used for fitting, ii) the raw spatial statistics results, iii) the designmatrix of the inference and iv) the fitted pffr object v) the residual standard error per condition defined as the residual sum of squares divided by the number of datapoints - sum of the estimated degrees of freedom for the model parameters as well as other QC metrics
adj.pvalue	a pvalue adjustment method as passed to stats::p.adjust defaults to Benjamini-Hochberg correction of the false discovery rate.
coefficientsToPlot	list of which coefficients to plot in the heatmap defaults to NULL in which case all coefficients are plotted
QCThreshold	the threshold on the Quality control metric. Depends on the function used.
QCMetric	the metric to relate the quality of the fit too.
...	other parameters passed to ggplot2 functions

Value

a ggplot2 object

Examples

spe <- .loadExample()
#make the condition a factor variable
colData(spe)[["patient_stage"]] <- factor(colData(spe)[["patient_stage"]])
#relevel to have non-diabetic as the reference category
colData(spe)[["patient_stage"]] <- relevel(colData(spe)[["patient_stage"]],
"Non-diabetic")

selection <- c("acinar", "ductal")
resLs <- crossSpatialInference(spe, selection,
                      fun = "Gcross", marks = "cell_type",
                      rSeq = seq(0, 50, length.out = 50), correction = "rs",
                      sample_id = "patient_id",
                      image_id = "image_number", condition = "patient_stage",
                      algorithm = "bam",
                      ncores = 1
                  )
p <- plotCrossHeatmap(resLs, adj.pvalue = "BH")

---

Plot a cross type spatial metric per field of view

Description

This function plots the cross function between two marks output from calcMetricPerFov. It wraps around helper function and applies this function to all samples.

Usage

plotCrossMetricPerFov(
  metricDf,
  theo = NULL,
  correction = NULL,
  x = NULL,
  imageId = NULL,
  ID = NULL,
  nrow = NULL,
  ncol = NULL,
  legend.position = "none",
  ...
)

Arguments

metricDf	the metric dataframe as calculated by calcMetricPerFov
theo	logical; if the theoretical line should be plotted
correction	the border correction to plot
x	the x-axis variable to plot
imageId	the ID of the image/fov
ID	the (optional) ID for plotting combinations
nrow	the number of rows for the facet wrap
ncol	the number of columns for the facet wrap
legend.position	the position of the legend of the plot
...	Other parameters passed to ggplot2 functions

Value

a ggplot object

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcCrossMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)

metricRes$ID <- paste0(
 metricRes$patient_stage, "|", metricRes$patient_id
)

metricRes <- subset(metricRes, image_number %in% c(138, 139, 140))
p <- plotCrossMetricPerFov(metricRes,
    theo = TRUE, correction = "rs",
    x = "r", imageId = "image_number", ID = "ID"
)
print(p)

---

Functional boxplot of spatstat curves

Description

This function creates a functional boxplot of the spatial statistics curves. It creates one functional boxplot per aggregation category, e.g. condition.

Usage

plotFbPlot(metricDf, x, y, aggregateBy)

Arguments

metricDf	the metric dataframe as calculated by calcMetricPerFov
x	the name of the x-axis of the spatial metric
y	the name of the y-axis of the spatial metric
aggregateBy	the criterion by which to aggregate the curves into a functional boxplot. Can be e.g. the condition of the different samples.

Value

a list of base R plots

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)
# create a unique ID for the data preparation
metricRes$ID <- paste0(
  metricRes$patient_stage, "|", metricRes$patient_id,
  "|", metricRes$image_number
)

plotFbPlot(metricRes, 'r', 'rs', 'patient_stage')

---

Plot a biplot from an fPCA analysis

Description

A function that takes the output from the functionalPCA function and returns a ggplot object of the first two dimensions of the PCA as biplot.

Usage

plotFpca(data, res, colourby = NULL, labelby = NULL)

Arguments

data	a data object for functional data analysis containing at least the functional response $Y$.
res	the output from the fPCA calculation
colourby	the variable by which to colour the PCA plot by
labelby	the variable by which to label the PCA plot by

Value

a list with components of fpca.face

Examples

# load the pancreas dataset
library("tidyr")
library("stringr")
library("dplyr")
# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
# calculate the Gcross metric for alpha and beta cells
metricRes <- calcMetricPerFov(spe, c("alpha", "beta"),
    subsetby = "image_number", fun = "Gcross",
    marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
    c("patient_stage", "patient_id", "image_number"), ncores = 1
)
metricRes$ID <- paste0(
  metricRes$patient_stage, "|", metricRes$patient_id,
  "|", metricRes$image_number
)

# prepare data for FDA
dat <- prepData(metricRes, "r", "rs")

# drop rows with NA
dat <- dat |> drop_na()
# create meta info of the IDs
splitData <- str_split(dat$ID, "|")
dat$condition <- factor(sapply(splitData, function(x) x[1]))
dat$patient_id <- factor(sapply(splitData, function(x) x[2]))
dat$image_id <- factor(sapply(splitData, function(x) x[3]))
# calculate fPCA
mdl <- functionalPCA(
    data = dat, r = metricRes$r |> unique()
)
p <- plotFpca(
    data = dat, res = mdl, colourby = "condition",
    labelby = "patient_id"
)
print(p)

---

Plot a pffr model object

Description

A function that takes a pffr object as calculated in functionalGam and plots the functional coefficients. The functions are centered such that their expected value is zero. Therefore, the scalar intercept has to be added to the output with the argument shift in order to plot the coefficients in their original range.

Usage

plotMdl(mdl, predictor, shift = NULL)

Arguments

mdl	a pffr model object
predictor	predictor to plot
shift	the value by which to shift the centered functional intercept. this will most often be the constant intercept

Value

ggplot object of the functional estimate

Examples

library("tidyr")
library("stringr")
library("dplyr")
# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)
# create a unique ID for each row
metricRes$ID <- paste0(
    metricRes$patient_stage, "x", metricRes$patient_id,
    "x", metricRes$image_number
)

dat <- prepData(metricRes, "r", "rs", sample_id = "patient_id",
    image_id = "image_number", condition = "patient_stage")

#' # drop rows with NA
dat <- dat |> drop_na()

# create a designmatrix
condition <- dat$patient_stage
# relevel the condition - can set explicit contrasts here
condition <- relevel(condition, "Non-diabetic")
designmat <- model.matrix(~condition)
# colnames don't work with the '-' sign
colnames(designmat) <- c(
    "(Intercept)", "conditionLong_duration",
    "conditionOnset"
)
# fit the model
mdl <- functionalGam(
    data = dat, x = metricRes$r |> unique(),
    designmat = designmat, weights = dat$npoints,
    formula = formula(Y ~ conditionLong_duration +
        conditionOnset + s(patient_id, bs = "re")),
        algorithm = "bam"
)
summary(mdl, re.test = FALSE)
plotLs <- lapply(colnames(designmat), plotMdl,
    mdl = mdl,
    shift = mdl$coefficients[["(Intercept)"]]
)

---

Plot a spatial metric per field of view

Description

A function that plots the output of the function calcMetricPerFov. The plot contains one curve per FOV and makes subplots by samples.

Usage

plotMetricPerFov(
  metricDf,
  theo = FALSE,
  correction = NULL,
  x = NULL,
  imageId = NULL,
  ID = NULL,
  nrow = NULL,
  ncol = NULL,
  legend.position = "none",
  ...
)

Arguments

metricDf	the metric dataframe as calculated by calcMetricPerFov
theo	logical; if the theoretical line should be plotted
correction	the border correction to plot
x	the x-axis variable to plot
imageId	the ID of the image/fov
ID	the (optional) ID for plotting combinations
nrow	the number of rows for the facet wrap
ncol	the number of columns for the facet wrap
legend.position	the position of the legend of the plot
...	Other parameters passed to ggplot2 functions

Value

a ggplot object

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)
# ceate a unique plotting ID
metricRes$ID <- paste0(
 metricRes$patient_stage, "|", metricRes$patient_id
)

p <- plotMetricPerFov(metricRes,
    correction = "rs", x = "r",
    imageId = "image_number", ID = "ID"
)
print(p)

---

Prepare data from calcMetricRes to be in the right format for FDA

Description

Prepare data from calcMetricRes to be in the right format for FDA

Usage

prepData(metricRes, x, y, sample_id = NULL, image_id = NULL, condition = NULL)

Arguments

metricRes	a dataframe as calculated by calcMetricRes - requires the columns ID (unique identifier of each row)
x	the name of the x-axis of the spatial metric
y	the name of the y-axis of the spatial metric
sample_id	the spe colData variable to mark the sample
image_id	the spe colData variable to mark the image
condition	the spe colData variable to mark the condition

Value

returns a list with three entries, the unique ID, the functional response Y and the weights

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
metricRes <- calcMetricPerFov(spe, c("alpha", "Tc"),
    subsetby = "image_number", fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), by = c(
        "patient_stage", "patient_id",
        "image_number"
    ),
    ncores = 1
)

# create a unique ID for each row
metricRes$ID <- paste0(
    metricRes$patient_stage, "|", metricRes$patient_id,
    "|", metricRes$image_number
)
dat <- prepData(metricRes, "r", "rs", sample_id = "patient_id",
image_id = "image_number", condition = "patient_stage")

---

print the fPCA results

Description

this is a function that prints a summary of the fPCA result of class fpca

Usage

## S3 method for class 'fpca'
print(x, ...)

Arguments

x	the result of function functionalPCA
...	other parameters passed to base generic function print

Value

a formatted overview of the fPCA result

Examples

# load the pancreas dataset
library("tidyr")
library("stringr")
library("dplyr")
# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
# calculate the Gcross metric for alpha and beta cells
metricRes <- calcMetricPerFov(spe, c("alpha", "beta"),
    subsetby = "image_number", fun = "Gcross",
    marks = "cell_type", rSeq = seq(0, 50, length.out = 50),
    c("patient_stage", "patient_id", "image_number"), ncores = 1
)
metricRes$ID <- paste0(
  metricRes$patient_stage, "|", metricRes$patient_id,
  "|", metricRes$image_number
)
# prepare data for FDA
dat <- prepData(metricRes, "r", "rs")

# drop rows with NA
dat <- dat |> drop_na()

# create meta info of the IDs
splitData <- strsplit(dat$ID, "|", fixed = TRUE)
dat$condition <- factor(sapply(splitData, function(x) x[1]))
dat$patient_id <- factor(sapply(splitData, function(x) x[2]))
dat$image_id <- factor(sapply(splitData, function(x) x[3]))
# calculate fPCA
mdl <- functionalPCA(
    data = dat, r = metricRes$r |> unique()
)
mdl

---

Heuristic for the choice of rMax

Description

Heuristic for the choice of rMax

Usage

rMaxHeuristic(spe, subsetby, marks)

Arguments

spe	a SpatialExperiment object
subsetby	the spe colData variable to subset the data by. This variable has to be provided, even if there is only one sample.
marks	the marks to consider e.g. cell types

Value

a ggplot histogram of the bounding radius of all the

Examples

# retrieve example data from Damond et al. (2019)
spe <- .loadExample()
p <- rMaxHeuristic(spe,
subsetby = "image_number", marks = "cell_type"
)

---

Statistical Inference on Spatial Statistics Functions

Description

A function to perform spatial statistical inference on spatial omics data. This function works so far only on functions of radius "r".

Usage

spatialInference(
  spe,
  selection,
  fun,
  marks = NULL,
  rSeq = NULL,
  correction,
  sample_id,
  image_id,
  condition,
  continuous = FALSE,
  assay = "exprs",
  transformation = NULL,
  weights = "total",
  eps = 0.001,
  delta = "minNnDist",
  family = stats::gaussian(link = "log"),
  verbose = TRUE,
  upperDeltaProb = NULL,
  weightTransform = TRUE,
  AR1 = FALSE,
  sandwich = "cluster",
  ncores = 1,
  algorithm = "bam",
  ...
)

Arguments

spe	a SpatialExperiment object
selection	the mark(s) you want to compare. NOTE: This is directional. c(A,B) is not the same result as c(B,A).
fun	the spatstat function to compute on the point pattern object
marks	the marks to consider e.g. cell types
rSeq	the range of r values to compute the function over
correction	the edge correction to be applied
sample_id	the spe colData variable to mark the sample, if not NULL this will result in a mixed model estimation
image_id	the spe colData variable to mark the image
condition	the spe colData variable to mark the condition
continuous	A boolean indicating whether the marks are continuous defaults to FALSE
assay	the assay which is used if continuous = TRUE
transformation	the transformation to be applied as exponential e.g. 1/2 for sqrt or Fisher's variance-stabilising transformation if "Fisher"
weights	the weighting to be applied to the functional GAM. Either NULL (equal weights), total (npoints of total pattern), min (npoints of the smaller subpattern) or max (npoints of the larger subpattern) or a user defined value of same length as the number of curves to be estimated
eps	some distributional families fail if the response is zero, therefore, zeros can be replaced with a very small value eps
delta	the delta value to remove from the beginning of the spatial statistics functions. Can be reasonable if e.g. cells are always spaced by 10 µm. If set to "minNnDist" it will take the mean of the minimum nearest neighbour distance across all images for this cell type pair.
family	the distributional family for the functional GAM
verbose	logical indicating whether to print all information or not
upperDeltaProb	the quantile to filter out the constant 1 part for Gest and Gcross. If NULL no upper filtering is applied.
weightTransform	logical indicating whether the weights (number of points) should be sqrt transformed
AR1	logical indicating whether to calculate the autocorrelation of the residuals along the domain and account for this in a second fitting step
sandwich	string indicating how and if to adjust for heterscedasticity of the residuals with a sandwich correction
ncores	the number of cores to use for parallel processing, default = 1
algorithm	algorithm to fit the refund::pffr method. defaults to bam
...	Other parameters passed to spatstat.explore functions for parameters concerning the spatial function calculation and to refund::pffr for the functional additive mixed model inference

Value

a list with four objects: i) the dataframe with the spatial statistics results transformed and filtered as used for fitting, ii) the raw spatial statistics results, iii) the designmatrix of the inference and iv) the fitted pffr object v) the residual standard error per condition defined as the residual sum of squares divided by the number of datapoints - sum of the estimated degrees of freedom for the model parameters as well as other QC metrics

Examples

spe <- .loadExample()
#make the condition a factor variable
colData(spe)[["patient_stage"]] <- factor(colData(spe)[["patient_stage"]])
#relevel to have non-diabetic as the reference category
colData(spe)[["patient_stage"]] <- relevel(colData(spe)[["patient_stage"]],
"Non-diabetic")
res <- spatialInference(spe, c("alpha", "Tc"),
    fun = "Gcross", marks = "cell_type",
    rSeq = seq(0, 50, length.out = 50), correction = "rs",
    sample_id = "patient_id",
    image_id = "image_number", condition = "patient_stage",
    ncores = 1,
    algorithm = "bam"
)

---

Summary for functionalGam object

Description

Summary for functionalGam object

Usage

## S3 method for class 'functionalGam'
summary(object, ...)

Arguments

object	a fitted functionalGam-object
...	see summary.gam() for options.

Value

A list with summary information, see summary.gam()

Author(s)

Martin Emons, adapted from summary.pffr() by Fabian Scheipl

---