Skip to contents

Plot density-based tissue compartments alongside the density difference

Source: R/plot_tissue_split.R
plot_tissue_split.Rd

split_tissue() discards the density images and point patterns it computes the boundary from, to avoid inflating the mif. plot_tissue_split() recomputes them on demand, at plot time, from the settings recorded by split_tissue(), and draws the stored boundary polyline (never a recontoured one) on top of both a raster of the density difference and a scatter of the per-cell compartment label.

Usage

plot_tissue_split(
  mif,
  which = NULL,
  compartment = c("refined_density_compartment", "density_compartment"),
  panels = c("both", "density", "compartment"),
  colors = NULL,
  point_size = 0.4,
  raster_max_pixels = 5e+05,
  workers = 1,
  filename = NULL,
  path = NULL,
  settings = NULL,
  ...
)

Arguments

mif

object of class mif that has been through split_tissue()

which

samples to plot: NULL for all, or a numeric/character vector indexing/naming elements of mif$spatial

compartment

which compartment column to colour cells by: the 3-level "refined_density_compartment" (default) or the 2-level "density_compartment"

panels

which panel(s) to draw: "both" (default, one faceted plot with the density-difference raster and the compartment scatter side by side), "density" alone, or "compartment" alone

colors

named character vector of colours for class1, class2 and (for compartment = "refined_density_compartment") "Interface". If NULL, defaults to RColorBrewer::brewer.pal(3, "Set1")[1:2] for the two classes and "grey20" for "Interface"; NA/unclassified cells are always drawn "grey70".

point_size

size passed to ggplot2::geom_point() for the compartment panel

raster_max_pixels

the density raster is recomputed at plot time (see Details) at up to this many pixels; above the cap resolution is capped for display only – the drawn boundary line is unaffected and stays exact.

workers

number of cores to use, one sample per core

filename, path

if filename is given, every requested plot is also written to a single PDF at file.path(path, filename) (path defaults to the working directory; a trailing .pdf in filename is not doubled).

settings

provenance normally read from attr(mif$derived$density_boundary, "call_info"), which split_tissue() attaches. That attribute does not survive [ subsetting or dplyr::bind_rows() (e.g. after merge_mifs() disagrees across mifs), so pass the settings list explicitly to recover from that: a list with classifier, class1, class2, sigma, eps, interface_width, xloc, yloc, filter_density, rescale, min_density, hard_threshold, sample_id and spatialTIME_version. rescale, min_density and hard_threshold may be omitted: they are backfilled to the behaviour that predated them (FALSE, off, FALSE), so provenance recorded before those arguments existed still replots correctly.

...

accepts no arguments; present only so that a mistyped named argument above produces an informative "Unknown argument" error instead of being silently absorbed.

Value

a named list of ggplot objects, one per requested sample, named like mif$spatial – not the mif. Unlike plot_immunoflo(), this deliberately does not attach the plots to mif$derived: each plot captures a raster data frame that can be as large as the sample itself, and merge_mifs()'s handling of list-valued derived slots is fragile enough already (see NEWS.md) without doubling the exposure.

Examples

library(dplyr)
x <- create_mif(clinical_data = example_clinical %>%
  mutate(deidentified_id = as.character(deidentified_id)),
  sample_data = example_summary %>%
  mutate(deidentified_id = as.character(deidentified_id)),
  spatial_list = example_spatial[1],
  patient_id = "deidentified_id",
  sample_id = "deidentified_sample")
x <- split_tissue(x, classifier = "Classifier.Label",
  class1 = "Tumor", class2 = "Stroma",
  sigma = 40, interface_width = 100)
plots <- plot_tissue_split(x)