Single-Cell RNA-seq Preprocessing Pipeline

A unified interface for standardized single-cell RNA-seq preprocessing. It accepts raw counts (matrix/data.frame), AnnData (via anndata or anndataR), or existing Seurat objects. The analysis flow is fully customizable via a character string pipeline and a configuration list params.

Usage

SCPreProcess(sc, ...)

# Default S3 method
SCPreProcess(
  sc = NULL,
  ...,
  pipeline = "onsvpcetu",
  params = list(o = list(project = "SC_Screen_Proj", min.cells = 400L), n = list(), s =
    list(), v = list(), p = list(), e = list(), c = list(resolution = 0.6), t = list(), u
    = list()),
  quality_control = list(pattern = c("^MT-")),
  data_filter = list(nFeature_thresh = c(200L, 6000L), nCount_thresh = c(500L, 50000L),
    percent.mt = 20L, percent.rp = 60L),
  column2only_tumor = NULL
)

# S3 method for class 'R6'
SCPreProcess(
  sc,
  ...,
  pipeline = "onsvpcetu",
  params = list(o = list(project = "SC_Screen_Proj", min.cells = 400L), n = list(), s =
    list(), v = list(), p = list(), e = list(), c = list(resolution = 0.6), t = list(), u
    = list()),
  quality_control = list(pattern = c("^MT-")),
  data_filter = list(nFeature_thresh = c(200L, 6000L), nCount_thresh = c(500L, 50000L),
    percent.mt = 20L, percent.rp = 60L),
  column2only_tumor = NULL
)

# S3 method for class 'Seurat'
SCPreProcess(sc, column2only_tumor = NULL, ...)

Arguments

sc

Input data. Can be:

• Matrix/Data frame: Raw count matrix (genes x cells).

• AnnData: Python AnnData object (read via anndata or anndataR packages).

• Seurat: A Seurat object (automatically validated and repaired if necessary).

...

Additional arguments for backward compatibility (mapped to params) or verbose control.

pipeline

A character string defining the processing steps and order. Characters map to Seurat functions:

• 'o': CreateSeuratObject (Must be the first step and cannot be deleted)

• 'n': NormalizeData

• 's': ScaleData

• 'v': FindVariableFeatures

• 'p': RunPCA

• 'e': FindNeighbors (Because "n" is used)

• 'c': FindClusters

• 't': RunTSNE

• 'u': RunUMAP

• 'r': SCTransform (Alternative to n/s/v)

Default is "onsvpcetu".

params

A named list of lists containing arguments for each pipeline step. Keys match the pipeline characters (e.g., params$n for NormalizeData). Default structure:

list(
  o = list(project = "SC_Screen_Proj", min.cells = 400L), # do not pass `counts`
  n = list(),             # NormalizeData args
  s = list(),             # ScaleData args
  v = list(),             # FindVariableFeatures args
  c = list(resolution = 0.6),
  ...
)

quality_control

A list containing regex patterns for QC metric calculation. (See QCPatternDetect) Default: list(pattern = "^MT-"). Detected metrics (e.g., percent.mt) are added to meta.data.

data_filter

A list of thresholds for cell filtering. Default: assay ("RNA"), nFeature_RNA (200-6000), nCount_RNA (500-50000), percent.mt (<20), percent.rp (<60). Only metrics detected via quality_control are filtered, i.e., nFeature_RNA, nCount_RNA and percent.mt.

column2only_tumor

Optional character. Column name in metadata to filter for tumor cells (matches "Tumor", "Cancer", "Malignant", etc.). If NULL, no filtering is performed.

Value

A processed Seurat object with reductions, clusters, and QC metrics.

Details

Pipeline Strategy: The function parses the pipeline string and executes corresponding Seurat functions in order. To use SCTransform, simply change the pipeline string (e.g., "orpetu") and provide parameters in params$r.

Quality Control & Filtering: QC metrics are generated based on regex patterns in quality_control. Cells are then filtered based on thresholds in data_filter. Column names for filtering are auto-generated (e.g., pattern "^MT-" -> filter "percent.mt"). If confused about the column name, use SigBridgeR:::Pattern2Colname().

See also

Other single_cell_preprocess: FilterTumorCell(), FindRobustElbow(), Pattern2Colname(), QCPatternDetect(), RegisterSeuratMethod(), SCAnnotate(), SCIntegrate(), SCPreProcessStrategy, compatible_with_3.0.2()

Other input_preprocess: BulkPreProcess(), PhenoMap(), PhenoPreProcess()

Examples

if (FALSE) { # \dontrun{
# 1. Standard pipeline (LogNormalize -> Scale -> PCA -> UMAP)
obj <- SCPreProcess(
  sc = counts_matrix,
  pipeline = "onsvpcetu",
  params = list(c = list(resolution = 0.8))
)

# 2. SCTransform pipeline
obj_sct <- SCPreProcess(
  sc = counts_matrix,
  pipeline = "orpcu", # Create -> SCT -> PCA -> Clusters -> UMAP
  quality_control = list(pattern = c("^MT-", "^RP[LS]")),
  params = list(
    r = list(vars.to.regress = "percent.mt")
  )
)

# 3. Start from AnnData with tumor filtering
adata_object <- anndataR::read_h5ad("data.h5ad")
obj_ad <- SCPreProcess(
  sc = adata_object,
  column2only_tumor = "tissue"
)
} # }