Parallelize 'riskRegression' functions

Henrik Bengtsson

+ =

The futurize package allows you to easily turn sequential code into parallel code by piping the sequential code to the futurize() function. Easy!

TL;DR

library(futurize)
plan(multisession)
library(riskRegression)
library(survival)

set.seed(42)
d <- sampleData(200, outcome = "competing.risks")
fit <- CSC(Hist(time, event) ~ X1 + X2 + X7 + X8, data = d)
sc <- Score(list("CSC" = fit), data = d,
            formula = Hist(time, event) ~ 1,
            times = 5, B = 100, split.method = "bootcv") |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize riskRegression functions such as Score().

The riskRegression package provides tools for risk regression modeling and prediction in survival analysis with competing risks. It supports fitting cause-specific Cox regression models, Fine-Gray regression, and absolute risk regression models. The Score() function performs bootstrap cross-validation for model evaluation, which is an excellent candidate for parallelization.

Example: Bootstrap cross-validation with Score()

The Score() function evaluates prediction models via bootstrap cross-validation with metrics such as time-dependent AUC and Brier scores:

library(riskRegression)
library(survival)

set.seed(42)
d <- sampleData(200, outcome = "competing.risks")
fit <- CSC(Hist(time, event) ~ X1 + X2 + X7 + X8, data = d)

## Bootstrap cross-validation with 100 bootstrap samples
sc <- Score(list("CSC" = fit), data = d,
            formula = Hist(time, event) ~ 1,
            times = 5, B = 100, split.method = "bootcv")

Here Score() evaluates sequentially, but we can easily make it evaluate in parallel by piping to futurize():

library(futurize)
library(riskRegression)
library(survival)

set.seed(42)
d <- sampleData(200, outcome = "competing.risks")
fit <- CSC(Hist(time, event) ~ X1 + X2 + X7 + X8, data = d)

sc <- Score(list("CSC" = fit), data = d,
            formula = Hist(time, event) ~ 1,
            times = 5, B = 100, split.method = "bootcv") |> futurize()

This will distribute the bootstrap samples across the available parallel workers, given that we have set up parallel workers, e.g.

plan(multisession)

The built-in multisession backend parallelizes on your local computer and works on all operating systems. There are other parallel backends to choose from, including alternatives to parallelize locally as well as distributed across remote machines, e.g.

plan(future.mirai::mirai_multisession)

and

plan(future.batchtools::batchtools_slurm)

Supported Functions

The following riskRegression functions are supported by futurize():

• Score() for ‘list’

Without futurize: Manual PSOCK cluster setup

For comparison, here is what it takes to parallelize Score() using the parallel and doParallel packages directly, without futurize:

library(riskRegression)
library(survival)
library(parallel)
library(doParallel)

set.seed(42)
d <- sampleData(200, outcome = "competing.risks")
fit <- CSC(Hist(time, event) ~ X1 + X2 + X7 + X8, data = d)

## Set up a PSOCK cluster and register it with foreach
ncpus <- 4L
cl <- makeCluster(ncpus)
registerDoParallel(cl)

## Bootstrap cross-validation in parallel via foreach
sc <- Score(list("CSC" = fit), data = d,
            formula = Hist(time, event) ~ 1,
            times = 5, B = 100, split.method = "bootcv",
            parallel = "as.registered")

## Tear down the cluster
stopCluster(cl)
registerDoSEQ()  ## reset foreach to sequential

This requires you to manually create a cluster, register it with doParallel, and remember to tear it down and reset the foreach backend when done. If you forget to call stopCluster(), or if your code errors out before reaching it, you leak background R processes. You also have to decide upfront how many CPUs to use and what cluster type to use. Switching to another parallel backend, e.g. a Slurm cluster, would require a completely different setup. With futurize, all of this is handled for you - just pipe to futurize() and control the backend with plan().