distTests: Statistical tests for intergenic distance data

In pgpmartin/GeneNeighborhood: Analyze the Neighborhood (Upstream/Downstream Neighbors) of Genes

Description

Statistical tests for intergenic distance data

Usage

distTests(GeneSetDistances, Universe = NULL, MedianResample = TRUE, R = 10000)

Arguments

GeneSetDistances	A tibble with intergenic distances for the different gene sets, as generated by the dist2Neighbors function.
Universe	Character string indicating which set should be considered as the universe (or control set)
MedianResample	Logical. Should the resample test of the median be performed (defaults to TRUE)
R	integer giving the number of resampling to perform for the resampling test (Default to 1e4)

Value

A tibble with the following columns:

• GeneSet. Name of the gene set.

• Orientation. Orientation of the neighbor (same or opposite strand).

• Side. Upstream or Downstream.

• KS.pvalue. p-value of Kolmogorov-Smirnof test

• Wilcox.pvalue. p-value of Wilcoxon rank sum test (or Mann-Whitney U test).

• Independ.pvalue. p-value of the independance test

if MedianResample is TRUE the tibble will also contain this additional column:

• Resample.pvalue. p-value from the resampling test.

DETAILS

The following tests are possible:

• Kolmogorov-Smirnov test. See ks.test. Although not adapted to integer values, it gives conservative p-values for large enough gene sets.

• Wilcoxon rank sum test (or Mann-Whithney U test). See wilcox.test.

• Independence test. See independence_test in the coin package

• resample. A test based on random resampling of the universe distances.

Author(s)

Pascal GP Martin

See Also

ks.test, wilcox.test, independence_test

Examples

#' ## Obtain gene neighborhood information:
  GeneNeighbors <- getGeneNeighborhood(Genegr)
## Get a (random) set of (100) genes:
  set.seed(123)
  randGenes <- sample(names(Genegr), 100)
## Create a set enriched for close upstream genes:
  GenePool <- GeneNeighbors[!is.na(GeneNeighbors$UpstreamDistance),]
  Proba <- (max(GenePool$UpstreamDistance)-GenePool$UpstreamDistance) /
             sum(max(GenePool$UpstreamDistance)-GenePool$UpstreamDistance)
  Proba <- (1/(GenePool$UpstreamDistance+1)) / sum(1/(GenePool$UpstreamDistance+1))
  CloseUpstream <- sample(GenePool$GeneName, size = 100, prob = Proba)
## Extract distances for this set of genes and for all genes :
  myGeneSets <- list("RandomGenes" = randGenes,
                     "CloseUpstream" = CloseUpstream,
                     "AllGenes" = GeneNeighbors$GeneName)
  distForGeneSets <- dist2Neighbors(GeneNeighbors,
                                    myGeneSets)
## Compare distances for genesets to a control set (here "AllGene")
## (using only 1K permutations fo speed purposes here, prefer using > 1e4)
  distTests(distForGeneSets,
            Universe="AllGenes",
            MedianResample = TRUE,
            R=1e3)

pgpmartin/GeneNeighborhood documentation built on Sept. 2, 2021, 6:37 a.m.