Parallelize 'GSVA' functions

The 'GSVA' logo + The 'futurize' hexlogo = The 'future' logo

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(GSVA)

param <- gsvaParam(expr, geneSets)
es <- gsva(param) |> futurize()

Deprecation Warning

As of GSVA (>= 2.7.0), futurization is no longer supported. An informative error message will produced, if attempted.

The reason for this is that GSVA has moved to using BiocParallel::bpiterate() internally. That call is not supported by the BiocParallel::DoparParam() backend, which is the only one that currently bridges Futureverse and BiocParallel. See https://github.com/futureverse/futurize/issues/14 for details and status updates.

Introduction

This vignette demonstrates how to use this approach to parallelize the GSVA functions.

The GSVA Bioconductor package implements gene set variation analysis, a non-parametric, unsupervised method for estimating variation of gene set enrichment through the samples of an expression data set. The main function gsva() computes enrichment scores for each gene set and sample, which can be parallelized across gene sets.

Example: Running gsva() in parallel

The gsva() function computes gene set enrichment scores using different methods depending on the parameter object passed to it:

library(GSVA)

# Create example data
set.seed(42)
n_genes <- 200L
n_samples <- 120L
expr <- matrix(rnorm(n_genes * n_samples), nrow = n_genes, ncol = n_samples)
rownames(expr) <- paste0("gene", seq_len(n_genes))
colnames(expr) <- paste0("sample", seq_len(n_samples))

geneSets <- list(
  geneSet1 = paste0("gene", sample(n_genes, 30L)),
  geneSet2 = paste0("gene", sample(n_genes, 50L)),
  geneSet3 = paste0("gene", sample(n_genes, 40L))
)

param <- gsvaParam(expr, geneSets)
es <- gsva(param)

Here gsva() runs sequentially, but we can easily make it run in parallel by piping to futurize():

library(futurize)

es <- gsva(param) |> futurize()

This will distribute the work 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)

Other enrichment methods

GSVA supports multiple enrichment methods through different parameter objects. All of them can be parallelized with futurize():

## ssGSEA method
es <- gsva(ssgseaParam(expr, geneSets)) |> futurize()

## PLAGE method
es <- gsva(plageParam(expr, geneSets)) |> futurize()

## Combined z-score method
es <- gsva(zscoreParam(expr, geneSets)) |> futurize()

Rank-based functions

As of GSVA (>= 2.7.5), the two-step gsvaRanks() + gsvaScores() pipeline is deprecated in favor of a three-step gsvaRowNorm() + gsvaColRanks() + gsvaColScores() pipeline. Each of these steps can be parallelized with futurize():

gsvarnorm <- gsvaRowNorm(param) |> futurize()
gsvaranks <- gsvaColRanks(gsvarnorm) |> futurize()
es <- gsvaColScores(gsvaranks, geneSets(param)) |> futurize()

Supported Functions

The following GSVA functions are supported by futurize():