Class "GOHyperGResult"

Description

This class represents the results of a test for overrepresentation of GO categories among genes in a selected gene set based upon the Hypergeometric distribution.

For details on extracting information from this object, be sure to read the accessor documentation in the Category package: HyperGResult-accessors .

Seealso

HyperGResult-accessors

Author

Seth Falcon

Defunct Functions in GOstats Package

Description

The functions or variables listed here are no longer part of GOstats as they are not needed (any more).

Usage

combGOGraph()
hyperGtable()
hyperG2Affy()
selectedGenes()
GOHyperG()
GOKEGGHyperG()
getGoGraph()

Details

combGOGraph was replaced by join . hyperGtable was replaced by summary . hyperG2Affy was replaced by probeSetSummary . GOLeaves was replaced by graph::leaves . selectedGenes was replaced by geneIdsByCategory . GOHyperG was replaced by hyperGTest . GOKEGGHyperG was replaced by hyperGTest . getGoGraph was replaced by GOGraph .

Tools for manipulating GO and microarrays.

Description

A set of tools for interacting with GO and microarray data. A variety of basic manipulation tools for graphs, hypothesis testing and other simple calculations.

Details

list(list("ll"), list(" ", "Package: ", list(), " GOstats", list(), " ", "Version: ", list(), " 1.7.4", list(), " ", "Date: ", list(), " 23-08-2006", list(), " ", "biocViews: ", list(), " Statistics, Annotation, GO, MultipleComparisons", list(), " ", "Depends: ", list(), " graph (>= 1.9.25), GO, annotate, RBGL, xtable, Biobase, ", "genefilter, multtest, Category (>= 1.3.7), methods", list(), " ", "Imports: ", list(), " methods, Category", list(), " ", "Suggests: ", list(), " hgu95av2.db (>= 1.6.0)",

list(), "

", "License: ", list(), " Artistic", list(), " "))

Index: list(" ", "ALL Acute Lymphoblastic Leukemia Data from the Ritz ", " Laboratory ", "GOstats-defunct Defunct Functions in GOstats Package ", "Ndists Distance matrices for the BCR/ABL and NEG ", " subgroups. ", "compCorrGraph A function to compute a correlation based graph ", " from Gene Expression Data ", "compGdist A function to compute the distance between ",

"                        pairs of nodes in a graph.

", "dropECode Drop GO labels for specified Evidence Codes ", "getEvidence Get the Evidence codes for a set of GO terms. ", "getGOTerm Functions to Access GO data. ", "getOntology Get GO terms for a specified ontology ", "hasGOannote Check for GO annotation ", "idx2dimnames Index to Dimnames ", "makeGOGraph Construct a GO Graph ", "notConn Find genes that are not connected to the ",

"                        others.

", "oneGOGraph Construct the GO graph given a set of leaves. ", "shortestPath Shortest Path Analysis ", "simLL Functions to compute similarities between GO ", " graphs and also between Entrez Gene IDs based on ", " their induced GO graphs. ", "triadCensus Triad Functions ")

Further information is available in the following vignettes: list(list("ll"), list(" ", list("GOstats"), " ", list(), " Using GOstats (source, pdf)", list(), " ", list("GOusage"), " ", list(), " Basic GO Usage (source, pdf)", list(), " ", list("GOvis"), " ", list(), " Visualizing Data Using GOstats (source, pdf)", list(), " "))

Author

R. Gentleman with contributions from S. Falcon

Maintainer: R. Gentleman rgentlem@fhcrc.org

Distance matrices for the BCR/ABL and NEG subgroups.

Description

These are precomputed distance matrices between all transcription factors selected. In the future they will be computed on the fly but currently that takes about 3 hours and so precomputed versions are supplied.

Format

These are both distance matrices.

Usage

data(Ndists)
  data(Bdists)

Examples

data(Ndists)
data(Bdists)

Class "OBOHyperGResult"

Description

This class represents the results of a test for overrepresentation of OBO categories among genes in a selected gene set based upon the Hypergeometric distribution.

For details on extracting information from this object, be sure to read the accessor documentation in the Category package: HyperGResult-accessors .

Seealso

HyperGResult-accessors

Author

Robert Castelo

A function to compute a correlation based graph from Gene Expression Data

Description

Given a set of gene expression data (an instance of the ExpressionSet class) this function computes a graph based on correlations between the probes.

Usage

compCorrGraph(eSet, k = 1, tau = 0.6)

Arguments

Details

Zhou et al. describe a method of computing a graph between probes (genes) based on estimated correlations between probes. This function implements some of their methods.

Pearson correlations between probes are computed and then these are raised to the power k . Any of the resulting estimates that are less than tau in absolute value are set to zero.

Value

An instance of the graph class. With edges and edge weights determined by applying the algorithm described previously.

Seealso

compGdist

Author

R. Gentleman

References

Zhou et al., Transitive functional annotation by shortest-path analysis of gene expression data.

Examples

## Create an ExpressionSet to work with
set.seed(123)
exprMat <- matrix(runif(50 * 5), nrow=50)
genData <- new("ExpressionSet", exprs=exprMat)

corrG = compCorrGraph(genData)

A function to compute the distance between pairs of nodes in a graph.

Description

Given a graph, g , and a set of nodes in the graph, whNodes , Dijkstra's shortest path algorithm is used to compute the distance between all pairs of nodes in whNodes .

Usage

compGdist(g, whNodes, verbose = FALSE)

Arguments

Details

This function can be quite slow, computation of the pairwise distances is not especially fast and if whNodes is long then there are many of them to compute.

Value

A matrix containing the pairwise distances. It might be worth making this an instance of the dist class at some point.

Seealso

compCorrGraph

Author

R. Gentleman

Examples

example(compCorrGraph)
compGdist(corrG, nodes(corrG)[1:5])

Hypergeometric Tests for GO term association

Description

Given a GOHyperGParams instance containing a set of unique Entrez Gene Identifiers, a microarray annotation data package name, and the GO ontology of interest, this function will compute Hypergeomtric p-values for over or under-representation of each GO term in the specified ontology among the GO annotations for the interesting genes. The computations can be done conditionally based on the structure of the GO graph.

Arguments

Details

When conditional(p) == TRUE , the hyperGTest function uses the structure of the GO graph to estimate for each term whether or not there is evidence beyond that which is provided by the term's children to call the term in question statistically overrepresented.

The algorithm conditions on all child terms that are themselves significant at the specified p-value, odds ratio, minimum or maximum gene set size cutoff. Given a subgraph of one of the three GO ontologies, or the ontology given in the OBOHyperGParams instance, the terms with no child categories are tested first. Next the nodes whose children have already been tested are tested. If any of a given node's children tested significant, the appropriate conditioning is performed.

Value

A GOHyperGResult or OBOHyperGResult instance.

Seealso

GOHyperGResult-class , %code{link[Category]{geneCategoryHyperGeoTest}}, geneGoHyperGeoTest , geneKeggHyperGeoTest

Author

Seth Falcon

References

FIXME

Index to Dimnames

Description

A function to map from integer offsets in an array to the corresponding values of the row and column names. There is probably a better way but I didn't find it.

Usage

idx2dimnames(x, idx)

Arguments

Value

A list with two components. If it is a LIST, use

*

Seealso

dimnames

Author

R. Gentleman

Examples

data(Ndists)
ltInf = is.finite(Ndists)
xx = idx2dimnames(Ndists, ltInf)

Construct a GO Graph

Description

The directed acyclic graph (DAG) based on finding the most specific terms for the supplied Entrez Gene IDs is constructed and returned. The constructuion is per GO ontology (there are three, MF, BP and CC) and once the most specific terms have been identified then all less specific terms are found (these are the parents of the terms) and then their parents and so on, until the root is encountered.

Usage

makeGOGraph(x, Ontology = "MF", removeRoot = TRUE, mapfun = NULL,
            chip = NULL)

Arguments

Details

For each supplied Entrez Gene identifier all the GO annotations (in the specified ontology) are found. The mapping is achieved in one of three ways:

• If mapfun is provided, it will be used to perform the needed lookups. In this case, chip will be ignored.

• If chip is provided and mapfun=NULL , then the needed lookups will be done based on the Entrez to GO mappings encapsulated in the specified annotation data package. This is the recommended usage.

• If mapfun and chip are NULL or missing, then the function will attempt to load the GO package (the environment-based package, distinct from GO.db). This package contains a legacy environment mapping Entrez IDs to GO IDs. If the GO package is not available, an error will be raised. Omitting both mapfun and chip is not recommended as it is not compatible with the DB-based annotation data packages.

The mappings are different for the different ontologies. Typically a GO indentifier is used only in one specific ontology.

The resulting structure is stored in a graph using the graph package, again from Bioconductor.

Value

An object that inherits from the graph class. The particular implementation is not specified.

Seealso

oneGOGraph

Author

R. Gentleman

References

The Gene Ontology Consortium

Examples

library("hgu95av2.db")
set.seed(321)
gN <- unique(sample(keys(hgu95av2.db, 'ENTREZID'), 4))
gg1 <- makeGOGraph(gN, "BP", chip="hgu95av2.db")

Find genes that are not connected to the others.

Description

A function that takes as input a distance matrix and finds those entries that are not connected to any others (ie. those with distance Inf .

Usage

notConn(dists)

Arguments

Details

It is a very naive implementation. It presumes that not connected entries are not connected to any other entries, and this might not be true. Using the connComp function from the graph package or the RBGL package might be a better approach.

Value

A vector of the names of the items that are not connected.

Seealso

connComp

Author

R. Gentleman

Examples

data(Ndists)
notConn(Ndists)

Construct the GO graph given a set of leaves.

Description

Given one or more GO identifiers (which indicate the leaves in the graph) and a set of mappings to the less specific sets of nodes this function will construct the graph that includes that node and all children down to the root node for the ontology.

Usage

oneGOGraph(x, dataenv)
GOGraph(x, dataenv)

Arguments

Details

For any set of GO identifiers (from a common ontology) we define the induced GO graph to be that graph, based on the DAG structure (child - parent) of the GO ontology of terms, which takes the most specific set of GO terms that apply (for that ontology) and then joins these to all less specific terms. These functions help construct such graphs.

Value

The induced GO graph (or NULL) for the given GO identifier.

Seealso

makeGOGraph

Author

R. Gentleman

Examples

library("GO.db")
g1 <- oneGOGraph("GO:0003680", GOMFPARENTS)
g2 <- oneGOGraph("GO:0003701", GOMFPARENTS)
g3 <- join(g1, g2)

g4 <- GOGraph(c("GO:0003680", "GO:0003701"), GOMFPARENTS)
if( require("Rgraphviz") && interactive() )
plot(g3)

Summarize Probe Sets Associated with a hyperGTest Result

Description

Given the result of a hyperGTest run (an instance of GOHyperGResult ), this function lists all Probe Set IDs associated with the selected Entrez IDs annotated at each significant GO term in the test result.

Usage

probeSetSummary(result, pvalue, categorySize, sigProbesets, ids = "ENTREZID")

Arguments

Details

Usually the goal of doing a Fisher's exact test on a set of significant probesets is to find pathways or cellular activities that are being perturbed in an experiment. After doing the test, one usually gets a list of significant GO terms, and the next logical step might be to determine which probesets contributed to the significance of a certain term.

Because the input for the Fisher's exact test consists of a vector of unique Entrez Gene IDs, and there may be multiple probesets that interrogate a particular transcript, the ouput for this function lists all of the probesets that map to each Entrez Gene ID, along with an indicator that shows which of the probesets were used as input.

The rationale for this is that one might not be able to assume a given probeset actually interrogates the intended transcript, so it might be useful to be able to check to see what other similar probesets are doing.

Because one of the first steps before running hyperGTest is to subset the input vectors of geneIds and universeGeneIds, any information about probeset IDs that interrogate the same gene transcript is lost. In order to recover this information, one can pass a vector of probeset IDs that were considered significant. This vector will then be used to indicate which of the probesets that map to a given GO term were significant in the original analysis.

Value

A list of data.frame . Each element of the list corresponds to one of the GO terms (the term is provides as the name of the element). Each data.frame has three columns: the Entrez Gene ID ( EntrezID ), the probe set ID ( ProbeSetID ), and a 0/1 indicator of whether the probe set ID was provided as part of the initial input ( selected )

Note that this 0/1 indicator will only be correct if the 'geneId' vector used to construct the GOHyperGParams object was a named vector (where the names are probeset IDs), or if a vector of 'sigProbesets' was passed to this function.

Author

S. Falcon and J. MacDonald

Examples

## Fake up some data
library("hgu95av2.db")
library("annotate")
prbs <- ls(hgu95av2GO)[1:300]
## Only those with GO ids
hasGO <- sapply(mget(prbs, hgu95av2GO), function(ids)
if(!is.na(ids) && length(ids) > 1) TRUE else FALSE)
prbs <- prbs[hasGO]
prbs <- getEG(prbs, "hgu95av2")
## remove duplicates, but keep named vector
prbs <- prbs[!duplicated(prbs)]
## do the same for universe
univ <- ls(hgu95av2GO)[1:5000]
hasUnivGO <- sapply(mget(univ, hgu95av2GO), function(ids)
if(!is.na(ids) && length(ids) > 1) TRUE else FALSE)
univ <- univ[hasUnivGO]
univ <- unique(getEG(univ, "hgu95av2"))

p <- new("GOHyperGParams", geneIds=prbs, universeGeneIds=univ,
ontology="BP", annotation="hgu95av2", conditional=TRUE)
## this part takes time...
if(interactive()){
hyp <- hyperGTest(p)
ps <- probeSetSummary(hyp, 0.05, 10)
}

Shortest Path Analysis

Description

The shortest path analysis was proposed by Zhou et. al. The basic computation is to find the shortest path in a supplied graph between two Entrez Gene IDs. Zhou et al claim that other genes annotated along that path are likely to have the same GO annotation as the two end points.

Usage

shortestPath(g, GOnode, mapfun=NULL, chip=NULL)

Arguments

Details

The algorithm implemented here is quite simple. All Entrez Gene identifiers that are annotated at the GO node of interest are obtained. Those that are found as nodes in the graph are retained and used for the computation. For every pair of nodes at the GO term the shortest path between them is computed using sp.between from the RBGL package.

There is a presumption that the graph is undirected . This restriction could probably be lifted if there was some reason for it - a patch would be gratefully accepted.

The mapping of GO node to Entrez ID is achieved in one of three ways:

• If mapfun is provided, it will be used to perform the needed lookups. In this case, chip will be ignored.

• If chip is provided and mapfun=NULL , then the needed lookups will be done based on the GO to Entrez mappings encapsulated in the specified annotation data package. This is the recommended usage.

• If mapfun and chip are NULL or missing, then the function will attempt to load the GO package (the environment-based package, distinct from GO.db). This package contains a legacy environment mapping GO IDs to Entrez IDs. If the GO package is not available, an error will be raised. Omitting both mapfun and chip is not recommended as it is not compatible with the DB-based annotation data packages.

Value

The return values is a list with the following components:

*

Seealso

sp.between

Author

R. Gentleman

References

Transitive functional annotation by shortest-path analysis of gene expression data, by X. Zhou and M-C J. Kao and W. H. Wong, PNAS, 2002

Examples

library("hgu95av2.db")
library("RBGL")

set.seed(321)
uniqun <- function(x) unique(unlist(x))

goid <- "GO:0005778"
egIds <- uniqun(mget(uniqun(hgu95av2GO2PROBE[[goid]]),
hgu95av2ENTREZID))

v1 <- randomGraph(egIds, 1:10, .3, weights=FALSE)
## Since v1 is random, it might be disconnected and we need a
## connected graph to guarantee the existence of a path.
c1 <- connComp(v1)
largestComp <- c1[[which.max(sapply(c1, length))]]
v2 <- subGraph(largestComp, v1)

a1 <- shortestPath(v2, goid, chip="hgu95av2.db")

Functions to compute similarities between GO graphs and also between Entrez Gene IDs based on their induced GO graphs.

Description

Both simUI and simLP compute a similarity measure between two GO graphs. For simLL , first the induced GO graph for each of its arguments is found and then these are passed to one of simUI or simLP .

Usage

simLL(ll1, ll2, Ontology = "MF", measure = "LP", dropCodes = NULL,
      mapfun = NULL, chip = NULL)
simUI(g1, g2)
simLP(g1, g2)

Arguments

Details

For each of ll1 and ll2 the set of most specific GO terms within the ontology specified ( Ontology ) that are not based on any excluded evidence code ( dropCodes ) are found. The mapping is achieved in one of three ways:

• If mapfun is provided, it will be used to perform the needed lookups. In this case, chip will be ignored.

• If chip is provided and mapfun=NULL , then the needed lookups will be done based on the Entrez to GO mappings encapsulated in the specified annotation data package. This is the recommended usage.

• If mapfun and chip are NULL or missing, then the function will attempt to load the GO package (the environment-based package, distinct from GO.db). This package contains a legacy environment mapping Entrez IDs to GO IDs. If the GO package is not available, an error will be raised. Omitting both mapfun and chip is not recommended as it is not compatible with the DB-based annotation data packages.

Next, the induced GO graphs are computed.

Finally these graphs are passed to one of simUI , (union intersection), or simLP (longest path). For simUI the distance is the size of the intersection of the node sets divided by the size of the union of the node sets. Large values indicate more similarity. These similarities are between 0 and 1.

For simLP the length of the longest path in the intersection graph of the two supplied graph. Again, large values indicate more similarity. Similarities are between 0 and the maximum leaf depth of the graph for the specified ontology.

Value

A list with:

If one of the supplied Gene IDs does not have any GO terms associated with it, in the selected ontology and with the selected evidence codes then NA is returned.

Seealso

makeGOGraph

Author

R. Gentleman

Examples

library("hgu95av2.db")
eg1 = c("9184", "3547")

bb = simLL(eg1[1], eg1[2], "BP", chip="hgu95av2.db")

Extraction and Plotting of GO Terms from a GOHyperGResult Object

Description

These functions extract and plot graph instances representing the relationships among GO terms tested using hyperGTest .

Usage

termGraphs(r, id = NULL, pvalue = NULL, use.terms = TRUE)
inducedTermGraph(r, id, children = TRUE, parents = TRUE)
plotGOTermGraph(g, r = NULL, add.counts = TRUE, max.nchar = 20,
                node.colors=c(sig="lightgray", not="white"),
                node.shape="plaintext", ...)

Arguments

Details

list(" ", "
", " ", list(list("termGraphs"), list("returns a list of ", list("graph"), " objects each ", " representing one of the connected components of the subgraph of ", " the GO ontology induced by selecting the specified GO IDs (if ", " ", list("id"), " is present) or by selecting the GO IDs that have a ", " p-value less that ", list("pvalue"), ". If ", list("use.terms"), " is ", " ", list("TRUE"), " the GO IDs will be translated into GO Term names and ",

"      attached to the nodes as node attributes (see ", list("nodeData"), ").

", " Edges in the graphs go from child (more specific) to parent (less ", " specific).")), " ", " ", " ", list(list("inducedTermGraph"), list("returns a ", list("graph"), " object representing ", " the GO graph induced by the terms specified by ", list("id"), ". The ", " ", list("children"), " and ", list("parent"), " arguments control whether direct ", " children and/or direct parents of the terms specified by ",

list("id"), "

", " are added to the graph (at least one of the two must be ", " ", list("TRUE"), ").")), " ", " ", " ", list(list("plotGOTermGraph"), list("Create a plot using Rgraphviz of a ", " ", list("graph"), " object as returned by either ", list("termGraphs"), " or ", " ", list("inducedTermGraph"), ". If a ", list("GOHyperGResult"), " object is ", " provided, then the nodes will be colored according to significance ", " (based on the result object's ",

list("pvalueCutoff"), ") and counts will

", " be added to show the size of the categories. ", " ")), " ", " ", " ")

Author

Seth Falcon

Triad Functions

Description

These functions provide some tools for finding triads in an undirected graph. A triad is a clique of size 3. The function triadCensus returns a list of all triads.

Usage

triadCensus(graph)
isTriad(x, y, z, elz, ely)
reduce2Degreek(graph, k)
enumPairs(iVec)

Arguments

Details

enumPairs takes a vector as input and returns a list of length choose(length(iVec),2)/2 containing all unordered pairs of elements.

isTriad takes three nodes as arguments. It is already known that x has edges to both y and z and we want to determine whether these are reciprocated. This is determined by examining elz for both x and y and then examining ely for both x and z .

reduce2Degreek is a function that takes an undirected graph as input and removes all nodes of degree less than k . This process is iterated until there are no nodes left (an error is thrown) or all nodes remaining have degree at least k . The resultant subgraph is returned. It is used here because to be in a triad all nodes must have degree 2 or more.

triadCensus makes use of the helper functions described above and finds all triads in the graph.

Value

A list where each element is a triple indicating the members of the triad. Order is not important and all triads are reported in alphabetic order.

Note

See the graph package, RBGL and Rgraphviz for more details and alternatives.

Author

R. Gentleman

Examples

##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,