Seurat v5.6-beta

In Seurat v5.6, we introduce updates to the core Seurat pipeline that improve performance on large datasets.

The functions included in this release are NormalizeData, FindVariableFeatures, ScaleData, SCTransform, RunPCA, FindNeighbors, FindClusters, RunUMAP, FindMarkers, and FindAllMarkers.

Most of the speedups in this release come from software engineering and moving core computations to C++, and multithreading provides substantial gains for some functions (notably, NormalizeData, RunPCA, and FindClusters).

Besides (optionally) defining the number of threads as we describe below, no additional setup or changes to workflows are needed to take advantage of the performance improvements in this update. A detailed writeup will accompany the forthcoming CRAN release.

We encourage users to post issues (as a new issue) and questions (in this thread), and appreciate any and all feedback. Thanks for using Seurat!

---

Install the v5.6-beta branch

remotes::install_github("satijalab/seurat@v5.6-beta")

We recommend restarting the R session following installation. packageVersion("Seurat") will show 5.5.1.9999.

Set up multithreading

With no setup, Seurat functions run single-threaded by default. To take advantage of multithreading, users can use setThreads() and set the number of threads to be used for parallelization.

Calling setThreads() or setThreads(NULL) will set the number of threads to (number of cores) - 1; for example:

# for example, on a machine w/ a 8-core CPU:
setThreads()
getThreads() # return the number of threads currently set
# > 7

Alternatively, provide the desired number of threads to be used.

setThreads(8)