SCP: Single Cell Pipeline

SCP provides a comprehensive set of tools for single cell data processing and downstream analysis.

The package includes the following facilities:

• Integrated single cell quality control methods.
• Pipelines embedded with multiple methods for normalization, feature reduction, and cell population identification (standard Seurat workflow).
• Pipelines embedded with multiple integration methods for scRNA-seq or scATAC-seq data, including Uncorrected, Seurat, scVI, MNN, fastMNN, Harmony, Scanorama, BBKNN, CSS, LIGER, Conos, ComBat.
• Multiple single cell downstream analyses such as identification of differential features, enrichment analysis, GSEA analysis, identification of dynamic features, PAGA, RNA velocity, Palantir, Monocle2, Monocle3, etc.
• Multiple methods for automatic annotation of single-cell data and methods for projection between single-cell datasets.
• High quality data visualization methods.
• Fast deployment of single-cell data into SCExplorer, a shiny app that provides an interactive visualization interface.

The functions in the SCP package are all developed around the Seurat object and compatible with other Seurat functions.

R version requirement

• R >= 4.1.0

Installation in the global R environment

You can install the latest version of SCP from GitHub with:

if (!require("devtools", quietly = TRUE)) {
  install.packages("devtools")
}
devtools::install_github("zhanghao-njmu/SCP")

Create SCP python environment

To run functions such as RunPAGA or RunSCVELO, SCP requires conda to create a separate python environment. The default environment name is "SCP_env". You can specify the environment name for SCP by setting options(SCP_env_name="new_name")

Now, you can run PrepareEnv() to create the python environment for SCP. If the conda binary is not found, it will automatically download and install miniconda.

SCP::PrepareEnv()

To force SCP to use a specific conda binary, it is recommended to set reticulate.conda_binary R option:

options(reticulate.conda_binary = "/path/to/conda")
SCP::PrepareEnv()

If the download of miniconda or pip packages is slow, you can specify the miniconda repo and PyPI mirror according to your network region.

SCP::PrepareEnv(
  miniconda_repo = "https://mirrors.bfsu.edu.cn/anaconda/miniconda",
  pip_options = "-i https://pypi.tuna.tsinghua.edu.cn/simple"
)

Available miniconda repositories:

• https://repo.anaconda.com/miniconda (default)

• http://mirrors.aliyun.com/anaconda/miniconda

• https://mirrors.bfsu.edu.cn/anaconda/miniconda

• https://mirrors.pku.edu.cn/anaconda/miniconda

• https://mirror.nju.edu.cn/anaconda/miniconda

• https://mirrors.sustech.edu.cn/anaconda/miniconda

• https://mirrors.xjtu.edu.cn/anaconda/miniconda

• https://mirrors.hit.edu.cn/anaconda/miniconda

Available PyPI mirrors:

• https://pypi.python.org/simple (default)

• https://mirrors.aliyun.com/pypi/simple

• https://pypi.tuna.tsinghua.edu.cn/simple

• https://mirrors.pku.edu.cn/pypi/simple

• https://mirror.nju.edu.cn/pypi/web/simple

• https://mirrors.sustech.edu.cn/pypi/simple

• https://mirrors.xjtu.edu.cn/pypi/simple

• https://mirrors.hit.edu.cn/pypi/web/simple

Installation in a isolated R environment using renv

If there are some R package version conflicts, or you do not want to change your current R environment, you can use the renv package to install SCP into an isolated R environment.

Create a isolated R environment

if (!require("renv", quietly = TRUE)) {
  install.packages("renv")
}
dir.create("~/SCP_env", recursive = TRUE) # It cannot be the home directory "~" !
renv::init(project = "~/SCP_env", bare = TRUE, restart = TRUE)

Install SCP package and create SCP python environment

renv::activate(project = "~/SCP_env")
renv::install("BiocManager")
renv::install("zhanghao-njmu/SCP", repos = BiocManager::repositories())
SCP::PrepareEnv()

When installing packages, network issues may cause the download to fail. Sometimes you need to provide a GitHub personal access token and restart the r session before downloading.

Activate SCP environment first before use

renv::activate(project = "~/SCP_env")

library(SCP)
data("pancreas_sub")
pancreas_sub <- RunPAGA(srt = pancreas_sub, group_by = "SubCellType", linear_reduction = "PCA", nonlinear_reduction = "UMAP")
ClassDimPlot(pancreas_sub, group.by = "SubCellType", reduction = "draw_graph_fr")

Save and restore the state of SCP environment

renv::snapshot(project = "~/SCP_env")
renv::restore(project = "~/SCP_env")

Example

Load the Data

The analysis is based on a subsetted version of mouse pancreas data.

library(SCP)
data("pancreas_sub")
print(pancreas_sub)
#> An object of class Seurat 
#> 47874 features across 1000 samples within 3 assays 
#> Active assay: RNA (15958 features, 0 variable features)
#>  2 other assays present: spliced, unspliced
#>  2 dimensional reductions calculated: PCA, UMAP

Data exploration:

ClassDimPlot(
  srt = pancreas_sub, group.by = c("CellType", "SubCellType"),
  reduction = "UMAP", theme_use = "theme_blank"
)

ClassDimPlot(
  srt = pancreas_sub, group.by = "SubCellType", stat.by = "Phase",
  reduction = "UMAP", theme_use = "theme_blank"
)

ExpDimPlot(
  srt = pancreas_sub, features = c("Sox9", "Neurog3", "Fev", "Rbp4"),
  reduction = "UMAP", theme_use = "theme_blank"
)

ExpDimPlot(
  srt = pancreas_sub, features = c("Ins1", "Gcg", "Sst", "Ghrl"),
  compare_features = TRUE, label = TRUE, label_insitu = TRUE,
  reduction = "UMAP", theme_use = "theme_blank"
)

ht <- GroupHeatmap(
  srt = pancreas_sub,
  features = c(
    "Sox9", "Anxa2", # Ductal
    "Neurog3", "Hes6", # EPs
    "Fev", "Neurod1", # Pre-endocrine
    "Rbp4", "Pyy", # Endocrine
    "Ins1", "Gcg", "Sst", "Ghrl" # Beta, Alpha, Delta, Epsilon
  ),
  group.by = c("CellType", "SubCellType"),
  heatmap_palette = "YlOrRd",
  cell_annotation = c("Phase", "G2M_score", "Neurod2"),
  cell_palette = c("Dark2", "Paired", "Paired"),
  show_row_names = TRUE, row_names_side = "left",
  add_dot = TRUE, add_reticle = TRUE
)
print(ht$plot)

CellQC

pancreas_sub <- RunCellQC(srt = pancreas_sub)
ClassDimPlot(srt = pancreas_sub, group.by = "CellQC", reduction = "UMAP")

ClassStatPlot(srt = pancreas_sub, stat.by = "CellQC", group.by = "CellType", label = TRUE)

ClassStatPlot(
  srt = pancreas_sub,
  stat.by = c(
    "db_qc", "outlier_qc", "umi_qc", "gene_qc",
    "mito_qc", "ribo_qc", "ribo_mito_ratio_qc", "species_qc"
  ),
  plot_type = "upset", stat_level = "Fail"
)

Standard pipeline in SCP

pancreas_sub <- Standard_SCP(srt = pancreas_sub)
ClassDimPlot(
  srt = pancreas_sub, group.by = c("CellType", "SubCellType"),
  reduction = "StandardUMAP2D", theme_use = "theme_blank"
)

ClassDimPlot3D(srt = pancreas_sub, group.by = "SubCellType")

ExpDimPlot3D(srt = pancreas_sub, features = c("Sox9", "Neurog3", "Fev", "Rbp4"))

Integration pipeline in SCP

Example data for integration is a subsetted version of panc8(eight human pancreas datasets)

data("panc8_sub")
panc8_sub <- Integration_SCP(srtMerge = panc8_sub, batch = "tech", integration_method = "Seurat")
ClassDimPlot(
  srt = panc8_sub, group.by = c("celltype", "tech"), reduction = "SeuratUMAP2D",
  title = "Seurat", theme_use = "theme_blank"
)

UMAP embeddings based on different integration methods in SCP:

Cell projection between single-cell datasets

panc8_rename <- RenameFeatures(srt = panc8_sub, newnames = make.unique(capitalize(rownames(panc8_sub), force_tolower = TRUE)), assays = "RNA")
srt_query <- RunKNNMap(srt_query = pancreas_sub, srt_ref = panc8_rename, ref_umap = "SeuratUMAP2D")
ProjectionPlot(
  srt_query = srt_query, srt_ref = panc8_rename,
  query_group = "SubCellType", ref_group = "celltype"
)

Cell annotation using bulk RNA-seq datasets

data("ref_scMCA")
pancreas_sub <- RunKNNPredict(srt_query = pancreas_sub, bulk_ref = ref_scMCA, filter_lowfreq = 20)
ClassDimPlot(srt = pancreas_sub, group.by = "KNNPredict_classification", reduction = "UMAP", label = TRUE)

Cell annotation using single-cell datasets

pancreas_sub <- RunKNNPredict(
  srt_query = pancreas_sub, srt_ref = panc8_rename,
  ref_group = "celltype", filter_lowfreq = 20
)
ClassDimPlot(srt = pancreas_sub, group.by = "KNNPredict_classification", reduction = "UMAP", label = TRUE)

pancreas_sub <- RunKNNPredict(
  srt_query = pancreas_sub, srt_ref = panc8_rename,
  query_group = "SubCellType", ref_group = "celltype",
  return_full_distance_matrix = TRUE
)
ClassDimPlot(srt = pancreas_sub, group.by = "KNNPredict_classification", reduction = "UMAP", label = TRUE)

ht <- CellCorHeatmap(
  srt_query = pancreas_sub, srt_ref = panc8_rename,
  query_group = "SubCellType", ref_group = "celltype",
  nlabel = 3, label_by = "row",
  show_row_names = TRUE, show_column_names = TRUE
)
print(ht$plot)

PAGA analysis

pancreas_sub <- RunPAGA(
  srt = pancreas_sub, group_by = "SubCellType",
  linear_reduction = "PCA", nonlinear_reduction = "UMAP"
)
PAGAPlot(srt = pancreas_sub, reduction = "UMAP", label = TRUE, label_insitu = TRUE, label_repel = TRUE)

Velocity analysis

pancreas_sub <- RunSCVELO(
  srt = pancreas_sub, group_by = "SubCellType",
  linear_reduction = "PCA", nonlinear_reduction = "UMAP"
)
VelocityPlot(srt = pancreas_sub, reduction = "UMAP", group_by = "SubCellType")

VelocityPlot(srt = pancreas_sub, reduction = "UMAP", plot_type = "stream")

Differential expression analysis

pancreas_sub <- RunDEtest(srt = pancreas_sub, group_by = "CellType", fc.threshold = 1, only.pos = FALSE)
VolcanoPlot(srt = pancreas_sub, group_by = "CellType")

DEGs <- pancreas_sub@tools$DEtest_CellType$AllMarkers_wilcox
DEGs <- DEGs[with(DEGs, avg_log2FC > 1 & p_val_adj < 0.05), ]
# Annotate features with transcription factors and surface proteins
pancreas_sub <- AnnotateFeatures(pancreas_sub, species = "Mus_musculus", db = c("TF", "SP"))
ht <- ExpHeatmap(
  srt = pancreas_sub, group.by = "CellType", features = DEGs$gene, feature_split = DEGs$group1,
  species = "Mus_musculus", db = c("GO_BP", "KEGG", "WikiPathway"), anno_terms = TRUE,
  feature_annotation = c("TF", "SP"), feature_palcolor = list(c("gold", "steelblue"), c("forestgreen")),
  height = 5, width = 4
)
print(ht$plot)

Enrichment analysis(over-representation)

pancreas_sub <- RunEnrichment(
  srt = pancreas_sub, group_by = "CellType", db = "GO_BP", species = "Mus_musculus",
  DE_threshold = "avg_log2FC > 1 & p_val_adj < 0.05"
)
EnrichmentPlot(
  srt = pancreas_sub, group_by = "CellType", group_use = c("Ductal", "Endocrine"),
  plot_type = "bar"
)

EnrichmentPlot(
  srt = pancreas_sub, group_by = "CellType", group_use = c("Ductal", "Endocrine"),
  plot_type = "wordcloud"
)

EnrichmentPlot(
  srt = pancreas_sub, group_by = "CellType", group_use = c("Ductal", "Endocrine"),
  plot_type = "wordcloud", word_type = "feature"
)

Enrichment analysis(GSEA)

pancreas_sub <- RunGSEA(
  srt = pancreas_sub, group_by = "CellType", db = "GO_BP", species = "Mus_musculus",
  DE_threshold = "p_val_adj < 0.05"
)
GSEAPlot(srt = pancreas_sub, group_by = "CellType", group_use = "Endocrine")

GSEAPlot(srt = pancreas_sub, group_by = "CellType", group_use = "Endocrine", geneSetID = "GO:0007186")

Trajectory inference

pancreas_sub <- RunSlingshot(srt = pancreas_sub, group.by = "SubCellType", reduction = "UMAP")

ExpDimPlot(pancreas_sub, features = paste0("Lineage", 1:3), reduction = "UMAP", theme_use = "theme_blank")

ClassDimPlot(pancreas_sub, group.by = "SubCellType", reduction = "UMAP", lineages = paste0("Lineage", 1:3), lineages_span = 0.1)

Dynamic features

pancreas_sub <- RunDynamicFeatures(srt = pancreas_sub, lineages = c("Lineage1", "Lineage2"), n_candidates = 200)
ht <- DynamicHeatmap(
  srt = pancreas_sub, lineages = c("Lineage1", "Lineage2"),
  use_fitted = TRUE, n_split = 6, reverse_ht = "Lineage1",
  species = "Mus_musculus", db = "GO_BP", anno_terms = TRUE, anno_keys = TRUE, anno_features = TRUE,
  heatmap_palette = "viridis", cell_annotation = "SubCellType",
  separate_annotation = list("SubCellType", c("Nnat", "Irx1")), separate_palette = c("Paired", "Set1"),
  feature_annotation = c("TF", "SP"), feature_palcolor = list(c("gold", "steelblue"), c("forestgreen")),
  pseudotime_label = 25, pseudotime_label_color = "red",
  height = 5, width = 2
)
print(ht$plot)

DynamicPlot(
  srt = pancreas_sub, lineages = c("Lineage1", "Lineage2"), group.by = "SubCellType",
  features = c("Plk1", "Hes1", "Neurod2", "Ghrl", "Gcg", "Ins2"),
  compare_lineages = TRUE, compare_features = FALSE
)

ExpStatPlot(
  srt = pancreas_sub, group.by = "SubCellType", bg.by = "CellType",
  features = c("Sox9", "Neurod2", "Isl1", "Rbp4"),
  comparisons = list(
    c("Ductal", "Ngn3 low EP"),
    c("Ngn3 high EP", "Pre-endocrine"),
    c("Alpha", "Beta")
  ),
  multiplegroup_comparisons = TRUE
)

Interactive data visualization with SCExplorer

PrepareSCExplorer(list(mouse_pancreas = pancreas_sub, human_pancreas = panc8_sub), base_dir = "./SCExplorer")
app <- RunSCExplorer(base_dir = "./SCExplorer")
list.files("./SCExplorer") # This directory can be used as site directory for Shiny Server.

if (interactive()) {
  shiny::runApp(app)
}

Other visualization examples

ClassDimPlot ClassStatPlot GroupHeatmap

More examples of SCP can be found in the documentation of the functions, such as Integration_SCP, RunKNNMap, RunMonocle3, RunPalantir, etc.