Advice on how to recordPlot() in future_lapply()

[leonfodoulian]

Hi,

I have a function that silently returns a plot that I would like to save in my output data using future.apply::future_lapply(). Any idea on how I could achieve this?

Thank you for your advice!

Best,
Leon

[scottkosty]

Hi, could you please make a minimal example? Are you using base graphics or ggplot2?

Also, it might help to know what you want to do with those plots later. Do you want to write them to disk? Or output them in another way? Or modify them before saving them to disk?

[leonfodoulian]

Hi @HenrikBengtsson,

Thank you for your answer. recordPlot() works using lapply(). I will give R.devices::capturePlot() a try and let you know.

Best,
Leon

[HenrikBengtsson]

Thank you for your answer. recordPlot() works using lapply()

Can you share a minimal example showing how that works and how you runit? Maybe that helps us to help you. It hard to say something useful with just a "hypothetical example".

[leonfodoulian]

Below is a replicate of my code using either lapply() or future.apply::future_lapply():

# Install packages
remotes::install_github(repo = 'tallulandrews/M3Drop')
remotes::install_github(repo = 'tallulandrews/M3DExampleData')

# Required packages
require(M3Drop)
require(M3DExampleData)

# Create a list of count data
counts.mat.l <- list("sample_1" = M3DExampleData::Mmus_example_list$data,
                     "sample_2" = M3DExampleData::Mmus_example_list$data)

# Identify variable genes using lapply
var.genes.lapply <- lapply(
  X = counts.mat.l,
  FUN = function(counts.mat) {
    # Remove all genes not expressed in at least 3 cells
    genes.to.keep <- rowSums(x = counts.mat > 0) >= 3 
    counts.mat <- counts.mat[genes.to.keep,]
    
    # Fit DANB to the raw counts matrix
    danb.fit <- M3Drop::NBumiFitModel(counts = counts.mat)
    
    # Check fit quality of DANB
    stats.danb <- M3Drop::NBumiCheckFitFS(counts = counts.mat,
                                          fit = danb.fit,
                                          suppress.plot = FALSE)
    fit.plot <- recordPlot()
    
    # Perform dropout-based DANB feature selection
    nb.drop.fs <- M3Drop::NBumiFeatureSelectionCombinedDrop(fit = danb.fit,
                                                            ntop = NULL,
                                                            method = "fdr",
                                                            qval.thresh = 0.01,
                                                            suppress.plot = FALSE)
    feature.selection.plot <- recordPlot()
    
    # Print "Done" to keep track of progress
    message("Done")
    
    # Return list of data
    return(list(danb.fit = danb.fit,
                stats.danb = stats.danb,
                nb.drop.fs = nb.drop.fs,
                plots = list(fit.plot = fit.plot,
                             feature.selection.plot = feature.selection.plot)))
  })

# Identify variable genes using future_lapply
future::plan(strategy = future::multisession)
var.genes.futurelapply <- future.apply::future_lapply(
  X = counts.mat.l,
  FUN = function(counts.mat) {
    # Remove all genes not expressed in at least 3 cells
    genes.to.keep <- rowSums(x = counts.mat > 0) >= 3 
    counts.mat <- counts.mat[genes.to.keep,]
    
    # Fit DANB to the raw counts matrix
    danb.fit <- M3Drop::NBumiFitModel(counts = counts.mat)
    
    # Check fit quality of DANB
    stats.danb <- M3Drop::NBumiCheckFitFS(counts = counts.mat,
                                          fit = danb.fit,
                                          suppress.plot = FALSE)
    fit.plot <- recordPlot()
    
    # Perform dropout-based DANB feature selection
    nb.drop.fs <- M3Drop::NBumiFeatureSelectionCombinedDrop(fit = danb.fit,
                                                            ntop = NULL,
                                                            method = "fdr",
                                                            qval.thresh = 0.01,
                                                            suppress.plot = FALSE)
    feature.selection.plot <- recordPlot()
    
    # Print "Done" to keep track of progress
    message("Done")
    
    # Return list of data
    return(list(danb.fit = danb.fit,
                stats.danb = stats.danb,
                nb.drop.fs = nb.drop.fs,
                plots = list(fit.plot = fit.plot,
                             feature.selection.plot = feature.selection.plot)))
  },
  future.seed = TRUE,
  future.scheduling = FALSE)
future::plan(strategy = future::sequential)

[HenrikBengtsson]

Can you please try to make a minimal example without all those packages? Something that just calls a basic plot function?

[leonfodoulian]

Sorry for the inconvenience. I hope that using the base iris dataset is fine.

# Function
my_function <- function(data) {
  plot(x = data$x,
       y = data$y)
  return(data$x * data$y)
}

# Input data
input.data.l <- list("data_1" = data.frame(x = iris$Sepal.Length,
                                           y = iris$Sepal.Width),
                     "data_2" = data.frame(x = iris$Petal.Length,
                                           y = iris$Petal.Width))

# Run function on list of input data using lapply
output.data.lapply <- lapply(
  X = input.data.l,
  FUN = function(input.data) {
    # Get output data from function
    output.data <- my_function(data = input.data)
    
    # Record plot to save with output data
    output.plot <- recordPlot()
    
    # Return list of data
    return(list(output.data = output.data,
                output.plot = output.plot))
  })

# Run function on list of input data using future_lapply
future::plan(strategy = future::multisession)
output.data.futurelapply <- future.apply::future_lapply(
  X = input.data.l,
  FUN = function(input.data) {
    # Get output data from function
    output.data <- my_function(data = input.data)
    
    # Record plot to save with output data
    output.plot <- recordPlot()
    
    # Return list of data
    return(list(output.data = output.data,
                output.plot = output.plot))
  },
  future.seed = TRUE,
  future.scheduling = FALSE)
future::plan(strategy = future::sequential)