compositePosteriorTest: Perform composite test on results from mashr

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

Perform composite test on results from mashr

Description

The posterior probabilities for all genes and conditions is obtained as 1-lFSR. Let prob be an array storing results for one gene. The probability that _no_ conditions in the exclusion set are non-zero is prod(1 - prob[exclude]). The probability that _all_ conditions in the inclusion set are non-zero is prod(prob[include]). The probability that _at least one_ condition in the inclusion set is non-zero is 1 - prod(1 - prob[include]). The composite test is the product of the probabilties computed from the inclusion and exclusion sets.

Usage

compositePosteriorTest(
  x,
  include,
  exclude = NULL,
  test = c("at least 1", "all")
)

Arguments

x	"dreamlet_mash_result" from run_mash()
include	array of conditions in the inclusion set
exclude	array of conditions in the exclusion set. Defaults to NULL for no exclusion
test	evaluate the posterior probability of a non-zero effect in "at least 1" or "all" conditions

Details

Perform composite test evaluating the specificity of an effect. Evalute the posterior probability that an a non-zero effect present in _all_ or _at least one_ condition in the inclusion set, but _no conditions_ in the exclusion set.

See Also

run_mash()

Examples

library(muscat)
library(mashr)
library(SingleCellExperiment)

data(example_sce)

# create pseudobulk for each sample and cell cluster
pb <- aggregateToPseudoBulk(example_sce[1:100, ],
  assay = "counts",
  cluster_id = "cluster_id",
  sample_id = "sample_id",
  verbose = FALSE
)

# voom-style normalization
res.proc <- processAssays(pb, ~group_id)

# Differential expression analysis within each assay,
# evaluated on the voom normalized data
res.dl <- dreamlet(res.proc, ~group_id)

# run MASH model
# This can take 10s of minutes on real data
# This small datasets should take ~30s
res_mash <- run_mash(res.dl, "group_idstim")

# Composite test based on posterior probabilities
# to identify effect present in *at least 1* monocyte type
# and *NO* T-cell type.
include <- c("CD14+ Monocytes", "FCGR3A+ Monocytes")
exclude <- c("CD4 T cells", "CD8 T cells")

# Perform composite test
prob <- compositePosteriorTest(res_mash, include, exclude)

# examine the lFSR for top gene
get_lfsr(res_mash$model)[which.max(prob), , drop = FALSE]

# Test if *all* cell types have non-zero effect
prob <- compositePosteriorTest(res_mash, assayNames(res.dl))

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