SCUBA is a unified data accession interface for single cell data formats. The package pulls data from Seurat, SingleCellExperiment, and Anndata objects for use in downstream plotting and analysis. Plotting functions that produce Seurat-style plots from all three object formats are also included.
SCUBA relies on the reticulate package to operate on AnnData objects. Please make sure you follow the instructions on the reticulate page before moving on.
It is recommended to use conda environments to manage your python version that is being used by reticulate. A conda environment can be used on the current R session by using reticulate::use_condaenv(). If you want to make it automatically load on startup you can select your conda environment in RStudio > Options > Python.
You must have the following packages installed and available in your python environment:
- Numpy
- Scipy
- Pandas
- AnnData
Run the command below to install SCUBA. BiocManager is used to automatically install Bioconductor dependencies (SCUBA is not a Bioconductor package).
If you plan to use SCUBA with anndata objects, use dependencies = TRUE. If you only plan to use SCUBA with Seurat and SingleCellExperiment objects, use dependencies = FALSE.
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
# Set dependencies to FALSE if you do not plan to use anndata objects
BiocManager::install("amc-heme/SCUBA", dependencies = TRUE)
The following demonstrates how to use SCUBA to access data in various formats.
There are three primary data accession methods: FetchData, fetch_metadata and fetch_reduction.
There are two object exploration methods: meta_varnames and unique_values
For Seurat objects
FetchData(
AML_Seurat,
layer = "data",
vars =
c("ab_CD117-AB",
"ab_CD123-AB",
"ab_CD11c-AB",
"rna_GAPDH",
"rna_MEIS1",
# Reductions
"UMAP_1",
"UMAP_2",
# Nonexistent features
"ab_CD900",
# Metadata
"nCount_RNA",
"nFeature_RNA",
# "Ambiguous" feature not in RNA assay
"CD11a-AB"
)
)
fetch_metadata(
AML_Seurat,
vars = c("Batch",
"nCount_RNA"
)
)
#Can use full_table=T which is much more computationally efficient than selecting vars.
fetch_reduction(
AML_Seurat,
dims = c(1, 2),
reduction = default_reduction(AML_Seurat),
cells = get_all_cells(AML_Seurat)
)
For SingleCellExperiment
FetchData(
AML_SCE(),
layer = "logcounts",
vars =
c("AB_CD117-AB",
"AB_CD123-AB",
"AB_CD11c-AB",
"RNA_GAPDH",
"RNA_MEIS1",
# Reductions
"UMAP_1",
"UMAP_2",
# Nonexistent features
"AB_CD900",
# Metadata
"nCount_RNA",
"nFeature_RNA",
# "Ambiguous" feature not in RNA assay
"CD11a-AB"
)
)
fetch_metadata(
AML_SCE(),
vars = c("Batch",
"nCount_RNA"
)
)
#Can use full_table=T which is much more computationally efficient than selecting vars.
fetch_reduction(
AML_SCE(),
dims = c(1, 2),
reduction = default_reduction(sce),
cells = get_all_cells(sce)
)
AnnData:
FetchData(
AML_h5ad(),
vars =
c("protein_CD123-AB",
"protein_CD117-AB",
"X_MEIS1",
"X_GAPDH",
# Reductions
"X_umap_1",
"X_umap_2",
# Metadata
"nCount_RNA",
"nFeature_RNA"
)
)
fetch_metadata(
AML_h5ad(),
vars = c("Batch",
"nCount_RNA"
)
)
#Can use full_table=T which is much more computationally efficient than selecting vars.
fetch_reduction(
AML_h5ad(),
dims = c(1, 2),
reduction = default_reduction(AML_anndata),
cells = get_all_cells(AML_anndata)
)
> meta_varnames(AML_h5ad())
[1] "nCount_RNA" "nFeature_RNA" "nCount_AB" "nFeature_AB" "nCount_BOTH" "nFeature_BOTH"
[7] "BOTH_snn_res.0.9" "seurat_clusters" "Prediction_Ind" "BOTH_snn_res.1" "ClusterID" "Batch"
[13] "x" "y" "x_mean" "y_mean" "cor" "ct"
[19] "prop" "meandist" "cDC" "B.cells" "Myelocytes" "Erythroid"
[25] "Megakaryocte" "Ident" "RNA_snn_res.0.4" "condensed_cell_type"
> unique_values(AML_h5ad(), var="condensed_cell_type")
[1] Plasma cells Primitive Dendritic cells Plasmacytoid dendritic cells
[5] BM Monocytes NK Cells CD8+ T Cells B Cells
[9] CD4+ T Cells PBMC Monocytes
10 Levels: B Cells BM Monocytes CD4+ T Cells CD8+ T Cells Dendritic cells NK Cells PBMC Monocytes ... Primitive
The following object exploration methods exist for Seurat and SingleCellExperiment objects. They are currently not implemented for AnnData objects.
reduction_names(AML_Seurat)
assay_names(AML_Seurat)
features_in_assay(AML_Seurat, "RNA")
SCUBA includes plotting functions that produce Seurat-style plots from all three object formats.
plot_feature(
AML_h5ad(),
feature = "UNG"
)
plot_dot(
AML_h5ad(),
group_by = "condensed_cell_type",
features = c("X_UNG", "X_GAPDH", "X_CCR5")
)
plot_reduction(
AML_h5ad(),
group_by="condensed_cell_type",
split_by="Batch"
)
plot_ridge(
AML_h5ad(),
group_by = "condensed_cell_type",
features = c("UNG", "GAPDH")
)
plot_scatter(
AML_h5ad(),
group_by = "condensed_cell_type",
feature_1 = "GAPDH",
feature_2 = "UNG"
)
plot_violin(
AML_h5ad(),
group_by = "condensed_cell_type",
features = c("UNG", "GAPDH")
)
Paper title and preprint link coming soon!
If any issues arise please use Github issues on this repository.
