Calculate Matrix Statistics Grouped by Clusters

Computes various statistics (mean, median, zscore, tscore, etc.) on a matrix, grouped by clusters.

CalcStats(object, ...)

# S3 method for Seurat
CalcStats(
  seu,
  features,
  group.by = NULL,
  cells = NULL,
  slot = "data",
  assay = NULL,
  method = "zscore",
  exp.transform = F,
  order = NULL,
  n = Inf,
  p.threshold = 0.05,
  priority = c("expr", "none")
)

# S3 method for default
CalcStats(
  matr,
  f,
  method = "zscore",
  exp.transform = F,
  t = F,
  order = NULL,
  n = Inf,
  p.threshold = 0.05
)

Arguments

object: An object. Can either be a Seurat object or a matrix.
seu: A Seurat object. Only applicable when using the Seurat method.
features: Features for computation, including gene expression, metrics, PC scores, or any other data retrievable via `FetchData()`. Defaults to NULL, implying all features in the matrix. Only applicable for the Seurat method.
group.by: A variable from `meta.data` for grouping or a character vector of equal length as the number of cells. Only applicable for the Seurat method.
cells: Cell identifiers to be used. Defaults to all cells. Only applicable for the Seurat method.
slot: Slot from which to fetch feature data. Only applicable for the Seurat method.
assay: Name of the assay to use. Defaults to the active assay. Only applicable for the Seurat method.
method: Computation method, one of: "mean", "median", "zscore", "tscore", "p", or "logFC". Default: "zscore".
exp.transform: Indicates if data should be transformed with `expm1`. Default: FALSE.
order: Sort rows by either "value" or "p" (t.test).
n: Top n rows for each cluster. Ignored if `order` is NULL. Default: Inf.
p.threshold: Rows with p-value greater than this threshold are omitted when `order = "p"`.
priority: If set to "expr", fetches data from the expression matrix over `meta.data`. Only applicable for the Seurat method.
matr: A matrix or data frame with rows as features and columns as cells.
f: A factor or vector indicating cell identity. It should match the column length of `matr`.
t: Transpose matrix if features are columns and cells are rows. Default: FALSE.

Value

A matrix.

Details

Computes statistics for each feature across cell types. "p" value is determined by t-test. For log-normalized data's LogFC computation, it's advised to set `exp.transform` to TRUE.

Examples

library(Seurat)
library(SeuratExtend)

# Using a Seurat object as input. First, select some genes for calculation:
genes <- VariableFeatures(pbmc)[1:20]

# Calculate zscore, grouping by the default 'ident' (cluster):
genes.zscore <- CalcStats(pbmc, features = genes, method = "zscore")
head(genes.zscore)

# Visualize with a heatmap:
Heatmap(genes.zscore, lab_fill = "zscore")

# Select more genes and retain the top 4 genes of each cluster, sorted by p-value.
# This can be a convenient method to display the top marker genes of each cluster:
genes <- VariableFeatures(pbmc)
genes.zscore <- CalcStats(
  pbmc, features = genes, method = "zscore", group.by = "cluster",
  order = "p", n = 4)
Heatmap(genes.zscore, lab_fill = "zscore")

# It's also possible to use a matrix as input. For instance, we initially perform
# Geneset Enrichment Analysis (GSEA) using the Hallmark 50 geneset and obtain the AUCell matrix
# (rows represent pathways, columns represent cells):
pbmc <- GeneSetAnalysis(pbmc, genesets = hall50$human)
matr <- pbmc@misc$AUCell$genesets

# Next, we calculate the zscore, grouped by 'cluster':
gsea.zscore <- CalcStats(matr, f = pbmc$cluster, method = "zscore")
Heatmap(gsea.zscore, lab_fill = "zscore")