vignettes/articles/Object_Conversion.Rmd
Object_Conversion.RmdWhile many packages are have some object converters they are not always as feature complete as desired. scCustomize provides a few straightforward converter functions.
FYI: Object converters can be fragile and/or not very flexible. I have tried to make these functions to avoid those issues. However, if that proves too hard a task to maintain long-term I may move them to separate package/deprecate them.
Load Seurat Object & Add QC Data
# read object
pbmc <- pbmc3k.SeuratData::pbmc3k.final
pbmc <- UpdateSeuratObject(pbmc)
pbmc <- Add_Cell_QC_Metrics(seurat_object = pbmc, species = "human")We’ll also add some random meta data variables to pbmc data form use in this vignette
pbmc$sample_id <- sample(c("sample1", "sample2", "sample3", "sample4", "sample5", "sample6"), size = ncol(pbmc),
replace = TRUE)scCustomize contains the conversion function as.LIGER().
This function has a few advantages over the conversion function
rliger::seuratToLiger().
as.liger will keep all meta data and transfer
it to LIGER by default.as.liger
will also store a reduction key for use with scCustomize LIGER plotting
functions to correctly set plot axes.as.liger can either take union or intersection
of genes between objects.as.liger has additional internal checks to ensure
compatibility with both Seurat V3/4 & V5 object structures.
pbmc_liger <- as.LIGER(x = pbmc, group.by = "sample_id")## • Checking Seurat object validity
## • Creating LIGER object.
## • Normalizing data.
pbmc_liger## An object of class liger
## with 6 datasets and
## 2638 total cells.Confirm that information from meta.data slot was transferred to LIGER object.
## [1] "nUMI" "nGene" "dataset"
## [4] "orig.ident" "seurat_annotations" "percent.mt"
## [7] "RNA_snn_res.0.5" "seurat_clusters" "percent_mito"
## [10] "percent_ribo"The liger/rliger package already contains a function
rliger::seuratToLiger() to convert LIGER objects to Seurat
Objects. However, during this transfer a few things have issues crop
up:
As of scCustomize v2.1.0 converting to Seurat objects from Liger can
be accomplished using as.liger function which functions
identically to previous function Liger_to_Seurat().
Liger_to_Seurat() will continue to work until v2.2.0 at
which point it will be completely deprecated.
scCustomize contains modified version of this function which extends
the Seurat function as.Seurat()
that solves these issues with some extra parameters:
keep_meta logical. Whether to keep meta data from the
@cell.data slot in LIGER object. Default is TRUE.reduction_label Name of dimensionality reduction
technique used (e.g., tSNE, UMAP, etc). Ensures dim names are set
correctly in Seurat object.seurat_assay Name of assay to use for data in Seurat
object. Default is “RNA”.assay_type Specify whether to create V3/4 vs V5 Seurat
assays.
new_seurat <- as.Seurat(x = pbmc_liger, reduction_label = "UMAP")## Preparing & merging matrices.
## Creating final sparse matrix.
## Normalizing layer: countsscCustomize also allows for the conversion of Seurat or LIGER objects
to python anndata objects for
analysis in scanpy or other
compatible python packages via the function as.anndata.
These functions were inspired/modified/updated from sceasy R package (see
as.anndata documentation).
as.anndata works with Seurat V3/4, Seurat V5, and LIGER
objects.as.anndata requires users have reticulate R
package and linked python installation with anndata installed.
as.anndata(x = pbmc, file_path = "~/Desktop", file_name = "pbmc_anndata.h5ad")## • Checking Seurat object validity & Extracting Data
## The following columns were removed as they contain identical values for all
## rows:
## ℹ orig.ident
## • Creating anndata object.
## • Writing anndata file: "/Users/marsh_mbp/Desktop/pbmc_anndata.h5ad"## AnnData object with n_obs × n_vars = 2638 × 13714
## obs: 'nCount_RNA', 'nFeature_RNA', 'seurat_annotations', 'percent.mt', 'RNA_snn_res.0.5', 'seurat_clusters', 'percent_mito', 'percent_ribo', 'percent_mito_ribo', 'log10GenesPerUMI', 'percent_top50', 'percent_oxphos', 'percent_apop', 'percent_dna_repair', 'percent_ieg', 'S.Score', 'G2M.Score', 'Phase', 'sample_id'
## var: 'vst.mean', 'vst.variance', 'vst.variance.expected', 'vst.variance.standardized', 'vst.variable'
## obsm: 'X_pca', 'X_umap'
## layers: 'counts'The release of Seurat V5+ has brought about two different types of assay structure that can exist within a Seurat object. However, some community tools that interact with Seurat objects have not been updated to work with both assay formats. Therefore it becomes necessary to change assay format for use with certain tools.
scCustomize provides Convert_Assay() for easy method to
convert from Assay>Assay5 (V3/4>5) or Assay5>Assay
(V5>V3/4).
# Convert to V5/Assay5 structure
pbmc_V5 <- Convert_Assay(seurat_object = pbmc, convert_to = "V5")
pbmc_V5[["RNA"]]## Assay (v5) data with 13714 features for 2638 cells
## Top 10 variable features:
## PPBP, LYZ, S100A9, IGLL5, GNLY, FTL, PF4, FTH1, GNG11, S100A8
## Layers:
## data, counts, scale.data
# Convert to V3/4/Assay structure
pbmc_V3 <- Convert_Assay(seurat_object = pbmc_V5, convert_to = "V3")
pbmc_V3[["RNA"]]## Assay data with 13714 features for 2638 cells
## Top 10 variable features:
## PPBP, LYZ, S100A9, IGLL5, GNLY, FTL, PF4, FTH1, GNG11, S100A8