aggregateVar: Per-sample variance of single-cell counts

In GabrielHoffman/dreamlet: Scalable differential expression analysis of single cell transcriptomics datasets with complex study designs

Per-sample variance of single-cell counts

Description

Aggregation function for single-cell log-normalized counts to calculate per-sample variance for dreamlet.

Usage

aggregateVar(
  sce,
  assay = NULL,
  cluster_id = NULL,
  sample_id = NULL,
  min.cells = 10,
  min.var = 0.01,
  min.samples = 4,
  min.prop = 0.4,
  verbose = TRUE,
  BPPARAM = SerialParam(progressbar = verbose)
)

Arguments

sce	a SingleCellExperiment.
assay	character string specifying the assay slot to use as input data. Defaults to the 1st available (assayNames(x)[1]).
cluster_id	character string specifying which variable to use as cluster id
sample_id	character string specifying which variable to use as sample id
min.cells	minimum number of observed cells for a sample to be included in the analysis
min.var	minimum variance for a gene to be considered expressed in a sample
min.samples	minimum number of samples passing cutoffs for cell cluster to be retained
min.prop	minimum proportion of retained samples with non-zero counts for a gene to be
verbose	logical. Should information on progress be reported?
BPPARAM	a BiocParallelParam object specifying how aggregation should be parallelized.

Details

The dreamlet workflow can also be applied to model gene expression variance. In this case, a per-sample per-gene variance is calculated across all cells from a given sample and cell type. Here aggregateVar() performs the roles of aggregateToPseudoBulk() followed by processAssays() but using log-normalized count data.

For each cell cluster, samples with at least min.cells are retained. Only clusters with at least min.samples retained samples are kept. Features are retained if they have at least min.var in at least min.prop fraction of the samples.

The precision of a measurement is the inverse of its sampling variance. The precision weights are computed as 1/sem^2, where sem = sd / sqrt(n) and n is the number of cells.

Value

a dreamletProcessedData object

Examples

library(muscat)
library(SingleCellExperiment)

data(example_sce)

# Compute variance for each sample and cell cluster
pbVar <- aggregateVar(example_sce,
  assay = "counts",
  cluster_id = "cluster_id",
  sample_id = "sample_id",
  verbose = FALSE
)

GabrielHoffman/dreamlet documentation built on Nov. 8, 2024, 2:45 a.m.