Expression Methods

DA.per

R Documentation

Permutation test of user-defined test statistic

Description

Test significance of multiple features with a binary predictor with a permutation scheme.

Usage

DA.per(
  data,
  predictor,
  paired = NULL,
  relative = TRUE,
  p.adj = "fdr",
  testStat = function(case, control) {     log2((mean(case) + 1)/(mean(control) + 1)) },
  testStat.pair = function(case, control) {     log2(mean((case + 1)/(control + 1))) },
  noOfIterations = 10000,
  margin = 50
)

Arguments

`data`	Either a matrix with counts/abundances, OR a `phyloseq` object. If a matrix/data.frame is provided rows should be taxa/genes/proteins and columns samples
`predictor`	The predictor of interest. Factor, OR if `data` is a `phyloseq` object the name of the variable in `sample_data(data)` in quotation
`paired`	For paired/blocked experimental designs. Either a Factor with Subject/Block ID for running paired/blocked analysis, OR if `data` is a `phyloseq` object the name of the variable in `sample_data(data)` in quotation
`relative`	Logical. Should `data` be normalized to relative abundances. Default TRUE
`p.adj`	Character. P-value adjustment. Default "fdr". See `p.adjust` for details
`testStat`	Function. Function for the test statistic. Should take two vectors as arguments. Default is a log fold change: `log2((mean(case abundances)+1)/(mean(control abundances)+1))`
`testStat.pair`	Function. Function for test statistic for paired analysis. Should take two vectors as arguments. Default is a log fold change: `log2(mean((case abundances+1)/(control abundances+1)))`
`noOfIterations`	Integer. Iterations for permutations. Default 10000
`margin`	Numeric. Margin for when to stop iterations if p-value is high and unlikely to become low

Details

Modified version of the one from: https://microbiomejournal.biomedcentral.com/articles/10.1186/s40168-016-0208-8. P-values are now two-sided, and test statistic is a simple log fold change

A paired permutation test is implemented specifically for this package. The test is similar to the original, but with a different test statistic and permutation scheme. The permutations are constrained in the paired version such that the predictor is only permuted within each level of the paired argument (e.g. subjects). The test statistic first finds the log-ratio between the two predictor levels (e.g. case and control) for each level of the paired argument and the final statistic is the mean of these log-ratios.

Value

A data.frame with with results.

Examples

# Creating random count_table and predictor
set.seed(4)
mat <- matrix(rnbinom(1000, size = 0.1, mu = 500), nrow = 100, ncol = 10)
rownames(mat) <- 1:100
pred <- c(rep("Control", 5), rep("Treatment", 5))

# Running permutation test on each feature
res <- DA.per(data = mat, predictor = pred)

Russel88/DAtest documentation built on March 24, 2022, 3:50 p.m.