combResults: Combining enrichment analysis results

View source: R/combine.R

combResultsR Documentation

Combining enrichment analysis results

Description

Different enrichment analysis methods usually result in different gene set rankings for the same dataset. This function allows to combine results from the different set-based and network-based enrichment analysis methods. This includes the computation of average gene set ranks across methods.

Usage

combResults(
  res.list,
  rank.col = configEBrowser("PVAL.COL"),
  decreasing = FALSE,
  rank.fun = c("comp.ranks", "rel.ranks", "abs.ranks"),
  comb.fun = c("mean", "median", "min", "max", "sum")
)

Arguments

res.list

A list of enrichment analysis result lists (as returned by the functions sbea and nbea).

rank.col

Rank column. Column name of the enrichment analysis result table that should be used to rank the gene sets. Defaults to the gene set p-value column, i.e. gene sets are ranked according to gene set significance.

decreasing

Logical. Should smaller (decreasing=FALSE, default) or larger (decreasing=TRUE) values in rank.col be ranked better? In case of gene set p-values the smaller the better, in case of gene set scores the larger the better.

rank.fun

Ranking function. Used to rank gene sets according to the result table of individual enrichment methods (as returned from the gsRanking function). This is typically done according to gene set p-values, but can also take into account gene set scores/statistics, especially in case of gene sets with equal p-value. Can be either one of the predefined functions ('comp.ranks', 'rel.ranks', 'abs.ranks') or a user-defined function. Defaults to 'comp.ranks', i.e. competitive (percentile) ranks are computed by calculating for each gene set the percentage of gene sets with a p-value as small or smaller. Alternatively, 'rel.ranks', i.e. relative ranks are computed in 2 steps:

  1. Ranks are assigned according to distinct gene set p-value *categories*, i.e. gene sets with equal p-value obtain the *same* rank. Thus, the gene sets with lowest p-value obtain rank 1, and so on.

  2. As opposed to absolute ranks (rank.fun = 'abs.ranks'), which are returned from step 1, relative ranks are then computed by dividing the absolute rank by number of distinct p-value categories and multiplying with 100 (= percentile rank).

comb.fun

Rank combination function. Used to combine gene set ranks across methods. Can be either one of the predefined functions (mean, median, max, min, sum) or a user-defined function. Defaults to 'sum', i.e. the rank sum across methods is computed.

Value

An enrichment analysis result list that can be detailedly explored by calling eaBrowse and from which a flat gene set ranking can be extracted by calling gsRanking.

Author(s)

Ludwig Geistlinger

See Also

sbea, nbea, eaBrowse

Examples



    # (1) expression data: 
    # simulated expression values of 100 genes
    # in two sample groups of 6 samples each
    se <- makeExampleData(what="SE")
    se <- deAna(se)

    # (2) gene sets:
    # draw 10 gene sets with 15-25 genes
    gs <- makeExampleData(what="gs", gnames=names(se))

    # (3) make artificial enrichment analysis results:
    # 2 ea methods with 5 significantly enriched gene sets each 
    ora.res <- makeExampleData(what="ea.res", method="ora", se=se, gs=gs) 
    gsea.res <- makeExampleData(what="ea.res", method="gsea", se=se, gs=gs)
    
    # (4) combining the results
    res.list <- list(ora.res, gsea.res)
    comb.res <- combResults(res.list)

    # (5) result visualization and exploration
    gsRanking(comb.res)

    # user-defined ranking and combination functions
    # (a) dummy ranking, give 1:nrow(res.tbl)
    dummy.rank <- function(res.tbl) seq_len(nrow(res.tbl))

    # (b) weighted average for combining ranks
    wavg <- function(r) mean(c(1,2) * r)

    comb.res <- combResults(res.list, rank.fun=dummy.rank, comb.fun=wavg)


lgeistlinger/EnrichmentBrowser documentation built on Oct. 29, 2023, 5:08 p.m.