In kevinrue/BioSeurat: Converter from Seurat to SingleCellExperiment

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Introduction

This package aims at making seurat objects (r CRANpkg("Seurat") package) accessible for downstream processing using common Bioconductor methods and classes. In particular, the most readily equivalent Bioconductor class apt to store the various components of a seurat object is the SingleCellExperiment class (r Biocpkg("SingleCellExperiment") package). Indeed, the SingleCellExperiment class contains slots to store: any number of assays, that encompasses the @raw.data, @data, and @scaled.data slots of a seurat object. sample meta-data, that corresponds to the @meta.data slot of a seurat object. feature meta-data, that encompasses the presence of individual genes in the @var.genes slot of a seurat object any number of reduced dimensions sets of coordinates, that are usually stored under the @dr slot of a seurat object

Pre-requisite

Installation

The r Githubpkg("kevinrue/BioSeurat") package is currently hosted on GitHub, and may be installed as follows:

devtools::install_github("kevinrue/BioSeurat", build_vignettes = TRUE)

Loading packages

Let us load the various packages required for this vignette;:

stopifnot(
    require(Seurat),
    require(SingleCellExperiment),
    require(BioSeurat)
)

Examples

Let us load both r CRANpkg("Seurat") and r Biocpkg("SingleCellExperiment") to set up the workspace

Minimal seurat object

Let us use the demonstration object of the r CRANpkg("Seurat") package:

pbmc_raw <- read.table(
  file = system.file('extdata', 'pbmc_raw.txt', package = 'Seurat'),
  as.is = TRUE
)
pbmc_small <- CreateSeuratObject(raw.data = pbmc_raw)
pbmc_small

This minimal seurat object can be converted to a SingleCellExperiment as follows:

pbmc_sce <- as(pbmc_small, "SingleCellExperiment")
pbmc_sce

Normalised seurat object

The NormalizeData function merely updates the @data slot of seurat objects (while also storing the associated parameters in the @calc.params slot):

pbmc_small <- NormalizeData(object = pbmc_small)
names([e-mail])

As a consequence, this seurat object can be converted to a SingleCellExperiment without additional difficulty:

pbmc_sce <- as(pbmc_small, "SingleCellExperiment")
pbmc_sce

Note: the normalisation parameters stored in the @calc.params slot are not transferred to the SingleCellExperiment object.

Scaled seurat object

The ScaleData function populates the @scale.data slot of seurat objects (while also storing the associated parameters in the @calc.params slot):

pbmc_small <- ScaleData(
    object = pbmc_small, vars.to.regress = c("nUMI"), display.progress = FALSE)
dim([e-mail])

The additional information in the @scale.data slot is stored as an additional assay of the same name in the resulting SingleCellExperiment object:

pbmc_sce <- as(pbmc_small, "SingleCellExperiment")
pbmc_sce

Variable genes

The FindVariableGenes function populates the @var.genes and @hvg.info slots of seurat objects (while also storing the associated parameters in the @calc.params slot):

pbmc_small <- FindVariableGenes(object = pbmc_small, do.plot = FALSE)
head([e-mail])

The information in the @hvg.info slot, as well as the presence of individual genes in the @var.genes slot are stored as columns in the @rowData slot of the resulting SingleCellExperiment object:

pbmc_sce <- as(pbmc_small, "SingleCellExperiment")
pbmc_sce

Coordinates in reduced dimensions

Functions such as RunPCA and RunTSNE add their respective coordinates under the @dr slot (while also storing the associated parameters in the @calc.params slot):

pbmc_small <- RunPCA(pbmc_small, do.print = FALSE)
names(pbmc_small@dr)
pbmc_small <-
    RunTSNE(pbmc_small, reduction.use = "pca", dims.use = 1:5, perplexity=10)
names(pbmc_small@dr)

Every cell embedding in the @dr slot will be transferred to the reducedDims slot of the resulting SingleCellExperiment object:

pbmc_sce <- as(pbmc_small, "SingleCellExperiment")
pbmc_sce

CCA-aligned

The RunCCA function combines two seurat objects, and stores the reduced dimensions obtained by canonical-correlation analysis (CCA) under the @dr slot (while also storing the associated parameters in the @calc.params slot).

pbmc1 <- SubsetData(pbmc_small,cells.use = [e-mail][1:40])
pbmc2 <- SubsetData(pbmc_small,cells.use = [e-mail][41:80])
[e-mail]$group <- "group1"
[e-mail]$group <- "group2"
pbmc_cca <- RunCCA(pbmc1, pbmc2)
pbmc_cca <- AlignSubspace(
    pbmc_cca,reduction.type = "cca", grouping.var = "group", dims.align = 1:2)
names(pbmc_cca@dr)

Naturally, PCA and t-SNE reduced dimensions from the original two seurat objects are not transferred to the new combined seurat objects. If desired, they need to be re-computed on the new object:

pbmc_cca <- RunPCA(pbmc_cca, do.print = FALSE)
names(pbmc_cca@dr)
pbmc_cca <-
    RunTSNE(pbmc_cca, reduction.use = "pca", dims.use = 1:5, perplexity=10)
names(pbmc_cca@dr)

Overall, CCA-aligned reduced dimensions are treated no differently than other dimensionality reduction embeddings; they are also transferred to the reducedDims slot of the resulting SingleCellExperiment object:

pbmc_sce <- as(pbmc_cca, "SingleCellExperiment")
pbmc_sce

Session info {.unnumbered}

sessionInfo()

kevinrue/BioSeurat documentation built on May 27, 2019, 2:09 p.m.