Skip to contents

Correlate one gene against all others within each cell type

CorrelateGene.Rd

Computes Spearman (or Pearson) correlation between a gene of interest (GOI) and every other gene, separately within each cell type. Three resolution levels are supported:

Usage

CorrelateGene(
  seurat_object,
  gene,
  group.by,
  level = c("sc", "pseudobulk", "group"),
  compare.by = NULL,
  sample.by = NULL,
  condition.by = NULL,
  assay = NULL,
  layer = "data",
  method = c("spearman", "pearson"),
  expressed_only = TRUE,
  min_pct = 10,
  min_cells = 20L,
  min_donors = 5L,
  min_groups = 3L
)

Arguments

seurat_object

A Seurat object.

gene

Character. The gene of interest (GOI).

group.by

Character. Cell-type identity column.

level

Character. "sc" (default), "pseudobulk", or "group".

compare.by

Character or NULL. Optional metadata column whose levels define independent comparison groups (e.g. "Sex", "Diagnosis"). When supplied, correlation is computed separately within each level so you can ask which genes are correlated with GOI in group A but not group B.

sample.by

Character. Donor/sample ID. Required for "pseudobulk".

condition.by

Character. Condition column. Required for "group".

assay

Character or NULL. Assay to use.

layer

Character. Layer/slot for SC and group levels. Default "data".

method

Character. "spearman" (default) or "pearson".

expressed_only

Logical. SC level: restrict to GOI > 0 cells? Default TRUE.

min_pct

Numeric. SC level: minimum % of (GOI-expressing) cells that must express a tested gene. Default 10.

min_cells

Integer. SC level: minimum cells per cell type. Default 20.

min_donors

Integer. Pseudobulk level: minimum donors. Default 5.

min_groups

Integer. Group level: minimum groups. Default 3.

Value

A data frame: gene, cell_type, r, p, padj, n, goi, level, method, and (when compare.by is set) compare_group. Sorted by cell type then descending |r|.

Details

  • "sc": cell level. Restricts to cells where GOI > 0 (expressed_only = TRUE) by default to avoid co-absence artifacts from dropout. Exploratory; validate hits at pseudobulk level.

  • "pseudobulk": donor level. Aggregates raw counts per donor x cell type, converts to edgeR log-CPM, then runs Spearman across donors. Statistically most valid.

  • "group": averages log-normalized expression within each condition.by x cell_type combination, then correlates across groups.

When compare.by is supplied, the analysis is run separately within each level of that variable (e.g. compare.by = "Sex" produces independent correlation tables for males and females). Results gain a compare_group column and PlotCorrelation() will lay the groups out side by side.

See also

SplitByGene(), PlotCorrelation()