Identify Pharmacological Cell Subpopulations — scPharmIdentify • scPharm

Classify single cells into drug-sensitive, drug-resistant, or other subpopulations based on pharmacogenomics profiles from the GDSC2 database.

Usage

scPharmIdentify(
  object,
  type,
  cancer,
  drug = NULL,
  nmcs = 50,
  nfeatures = 200,
  cores = 1,
  features = NULL,
  slot = "data",
  layer = NULL,
  assay = "RNA",
  threshold.s = -1.751302,
  threshold.r = 1.518551,
  tumor.cells = NULL,
  normal.cells = NULL,
  bulkdata = NULL,
  gdscdata = NULL
)

Arguments

object: A Seurat object containing single-cell RNA-seq data.
type: Data source type. Either "tissue" for tumor tissue samples or "cellline" for cell line samples. When "tissue", the function identifies tumor vs adjacent normal cells using CNV analysis.
cancer: TCGA cancer type(s). A character string or vector specifying cancer type(s) (e.g., "BRCA", c("LUAD", "LUSC")). Use "pan" for pan-cancer analysis.
drug: Drug name to analyze. If NULL (default), all drugs from GDSC2 project will be analyzed.
nmcs: Number of MCA components to compute. Default: 50.
nfeatures: Number of genes for cell identity signature. Default: 200.
cores: Number of CPU cores for parallel processing. Default: 1.
features: Character vector of gene names to use. If NULL, all features are used.
slot: Slot name for Seurat V4 data access. Default: "data".
layer: Layer name for Seurat V5 data access. If NULL, uses slot value.
assay: Assay to use. Default: "RNA".
threshold.s: Threshold for labeling sensitive cells. Cells with NES below this value are classified as sensitive. Default: -1.751302.
threshold.r: Threshold for labeling resistant cells. Cells with NES above this value are classified as resistant. Default: 1.518551.
tumor.cells: Character vector of known tumor cell barcodes. If provided when type="tissue", CNV-based detection is skipped.
normal.cells: Character vector of known normal cell barcodes. Used as reference for CNV analysis when provided.
bulkdata: Bulk RNA-seq data for cell lines. If NULL, uses built-in scPharm::bulkdata.
gdscdata: GDSC pharmacogenomics data. If NULL, uses built-in scPharm::gdscdata.

Value

A Seurat object with pharmacological annotations added to metadata:

cell.label: Cell type: "tumor" or "adjacent"
scPharm_label_DRUGID_DRUGNAME: Drug response: "sensitive", "resistant", or "other"
scPharm_nes_DRUGID_DRUGNAME: Normalized enrichment score for the drug

Details

The function performs the following steps:

For tissue samples: identifies tumor cells via CNV analysis (or uses provided annotations)
Computes Multiple Correspondence Analysis (MCA) for dimensionality reduction
Generates cell identity gene signatures
Correlates gene expression with drug response (AUC) across cell lines
Performs GSEA to score each cell's drug response profile
Classifies cells based on NES thresholds

References

Tian P, Zheng J, et al. scPharm: identifying pharmacological subpopulations of single cells for precision medicine in cancers. 2023.

Examples

if (FALSE) { # \dontrun{
# Basic usage
result <- scPharmIdentify(seurat_obj, type = "tissue", cancer = "LUAD")

# With known tumor cells
result <- scPharmIdentify(seurat_obj, type = "tissue", cancer = "LUAD",
                          tumor.cells = tumor_barcodes)

# Specific drug analysis
result <- scPharmIdentify(seurat_obj, type = "cellline", cancer = "BRCA",
                          drug = "Erlotinib")
} # }