Multiple testing on the GO graph {GlobalAncova}R Documentation

Multiple testing on the GO graph

Description

Three functions adapted from package globaltest to test (part of) the GO graph for association of the gene expression profile of GO terms with a response variable. Used together, these functions return multiplicity-adjusted p-values calculated using the Focus Level procedure that preserves the structure of the GO graph.

Usage

  GAGO(..., GO, focus, maxalpha = 0.05, stopafter = 100, verbose = FALSE)
  makeGOstructure(data, annotation, top, only.ids, ontology = c("BP", "CC", "MF"), unreliable)
  getFocus(GOstructure, maxatoms = 10)

Arguments

... Arguments describing the tests to be performed are passed on to GlobalAncova. Note that only the approximative version of GlobalAncova is used here and hence the parameter method is not available. Even though the number of permutations (perm) may be specified since very large gene sets (with more genes than max.group.size) are treated with the permutation test.
GO An object of class GOstructure describing the structure of the GO graph. This object should be created using makeGOstructure.
focus A vector of GO ids to describe te focus level. Typically made using getFocus.
maxalpha The maximum multiplicity-adjusted p-value. The algorithm will stop when this value is exceeded.
stopafter The maximum number of significant GO terms to be found. The algorithm will stop when this value is exceeded.
verbose If set to TRUE, prints much more extensive progress information.
data The data set for which the GOstructure object is to be made. Can be either an exprSet or a vector of probe indicators.
annotation The name of the metaData annotation package to be used, or a vector of EntrezGene ids with the same length as the number of genes in data.
top The node to be used as top node of the GOstructure. Defaults to one of "GO:0008150", "GO:0005575", "GO:0003674", depending on he ontology chosen. The GOstructure object will only contain offspring of the chosen top node.
only.ids A vector of GO ids. If this is supplied, GO ids not appearing in this list will not appear in the GOstructure object.
ontology The ontology to be used. One of Biological Process (BP), Cellular Component (CC) or Molecular Function (MF).
unreliable Can be used to designate one or more of the GO evidence codes as unreliable. Value must be a vector containing one or more from "IC", "IDA", "IEA", "IEP", "IGI", "IMP", "IPI", "ISS", "NAS", "ND", "RCA", "TAS", "NR".
GOstructure An object of class GOstructure.
maxatoms A tuning parameter that governs the choice of the focus level. getFocus will choose those GO terms in the focus level in such a way that all offspring gene sets of each focus level term can be constructed as unions of no more than maxatoms atom gene sets. Default value of maxatoms is 10. Higher values quickly lead to slower computation. Lower values typically lead to reduced power.

Details

These functions should be used in the following order. First use makeGOstructure to make a GO graph tailored to a specific data set. Then getFocus can be used to choose a focus level. Finally GAGO performs the focus level procedure.

Value

The function returns a named vector of multiplicity-adjusted p-values. Adjusted p-values of GO terms not appearing in this vector are larger than the chosen value of maxalpha.

Note

Function GAGO corresponds to function gtGO in package globaltest. The difference is in the use of the GlobalAncova test instead of globaltest within the focus level procedure.

Author(s)

Jelle Goeman: j.j.goeman@lumc.nl; Jan Oosting; Manuela Hummel

References

Goeman, J.J. and Mansmann, U., Family-wise error rate on the directed acyclic graph of Gene Ontology, submitted.

See Also

gtGO, GlobalAncova, globaltest, GOstructure

Examples

  # see vignettes of packages GlobalAncova and globaltest and help of gtGO

[Package GlobalAncova version 3.2.0 Index]