Run inferCNV copy-number inference

Wraps the full inferCNV workflow: gene-order table generation (via AnnotateFeatures() / GetGeneOrder()), downsampling, object creation, inference, post-processing of map_metadata_from_infercnv.txt, and standard figure output. Returns the Seurat object with new metadata columns: CNV_score, CNV_Prediction, infercnv_subcluster, and one proportion_cnv_chrN column per chromosome.

Usage

RunInferCNV(
  seurat_object,
  cell_type_col = "Assignment",
  ref_group_names,
  output_dir = NULL,
  species = NULL,
  chromosomes = NULL,
  gene_biotype = "protein_coding",
  annotation_file = NULL,
  downsample = 200L,
  cutoff = 0.1,
  HMM = FALSE,
  denoise = TRUE,
  scale_data = TRUE,
  cluster_by_groups = TRUE,
  num_threads = 4L,
  cnv_score_threshold = "auto",
  cnv_k = 2,
  resume = FALSE,
  caffeinate = FALSE
)

Arguments

seurat_object: A Seurat object.
cell_type_col: Character. Metadata column containing cell-type labels used to split cells and define reference groups. Default "Assignment".
ref_group_names: Character vector. Cell-type labels treated as the normal (diploid) reference — typically immune or stromal populations.
output_dir: Character or NULL. Directory for inferCNV output and saved figures. Walks up from PrepObject() settings when NULL.
species: Character. "mouse", "human", or "zebrafish". Used to auto-generate gene order via AnnotateFeatures() if not yet annotated.
chromosomes: Character or numeric vector of chromosomes to analyze. NULL uses species-appropriate autosomes (mouse 1–19, human 1–22, zebrafish 1–25).
gene_biotype: Character. Biotype filter for the gene order table. Default "protein_coding".
annotation_file: Character or NULL. Passed to AnnotateFeatures() when manual gene annotation is preferred over the automatic sources.
downsample: Integer. Maximum cells per identity class. NULL uses all cells (not recommended for large objects). Default 200L.
cutoff: Numeric. inferCNV expression cutoff. Use 0.1 for 10x Genomics and 1 for Smart-seq2. Default 0.1.
HMM: Logical. Run inferCNV's HMM for subclonal CNV calls? Dramatically increases runtime. Default FALSE.
denoise: Logical. Apply inferCNV denoising. Default TRUE.
scale_data: Logical. Scale expression data before inference. Default TRUE.
cluster_by_groups: Logical. Cluster cells within their annotation groups. Default TRUE.
num_threads: Integer. Threads for inferCNV. Default 4L.
cnv_score_threshold: Numeric or "auto". When "auto" (default), the threshold is set to mean(ref_scores) + cnv_k × sd(ref_scores) using the reference cells' CNV score distribution — this is the data-driven approach from Tirosh et al. 2016. Pass a numeric value to override.
cnv_k: Numeric. Number of standard deviations above the reference mean used for the auto threshold. Higher values = more conservative (fewer malignant calls). Default 2.
resume: Logical. If TRUE and output files already exist, skip re-running inferCNV and go straight to post-processing. Default FALSE.
caffeinate: Logical. Prevent the machine from sleeping during the run. Default FALSE.

Value

The input Seurat object with additional metadata columns.

Usage

Arguments

Value

See also