| Title: | This package is designed to perform Gene Set Analysis across multiple genomic platforms |
|---|---|
| Description: | the RTopper package is designed to perform and integrate gene set enrichment results across multiple genomic platforms. |
| Authors: | Luigi Marchionni <[email protected]>, Svitlana Tyekucheva <[email protected]> |
| Maintainer: | Luigi Marchionni <[email protected]> |
| License: | GPL (>= 3) + file LICENSE |
| Version: | 1.51.0 |
| Built: | 2024-07-13 05:04:29 UTC |
| Source: | https://github.com/bioc/RTopper |
Gene sets analysis considers whether genes sharing a biological
property also behave in a related way in experimental data.
This technique is commonly used in high throughput genomic
analyses to assist results interpretation, and has been successfully
applied in cancer genome projects for integrating information from
multiple genome-wide assays. The RTopper package uses
gene sets analysis to overcome the diversity of genomic data
providing the statistical framework for integration across data types.
| Package: | RTopper |
| Type: | Package |
| Version: | 0.1 |
| Date: | 2010-11-12 |
| License: | GPL version 3 or newer |
Rtopper enables two gene set-based data integration
approaches:
Integration+GSA: computing integrated gene-to-phenotype association scores, followed by conventional gene sets analysis;
GSA+Integration: computing consensus significance score after all data types are analyzed individually;
Use of alternative enrichment test: RTopper uses
the Wilcoxon rank-sum test for enrichment testing, however
alternative tests can be defined and used;
Multiple testing correction: RTopper enables
adjustment of p-values obtained from enrichment analysis;
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin, and Giovanni Parmigiani. "Integrating diverse genomic data using gene sets." Manuscript submitted.
This function is an interface to the mt.rawp2adjp function
contained in the multtest package.
adjustPvalGSE works on outputs from runBatchGSE
and combineGSE returning adjusted p-values
adjustPvalGSE(gseOut, proc = "BH", alpha = 0.05 , na.rm = FALSE)adjustPvalGSE(gseOut, proc = "BH", alpha = 0.05 , na.rm = FALSE)
gseOut |
list of lists, either the output from |
proc |
character, the method to be used for p-values adjusting.
This parameter will be passed to the |
alpha |
numeric, the nominal type I error rate |
na.rm |
logical, the option for handling NA values in the list of raw p-values |
The adjustPvalGSE function performs p-value adjusting
for multiple testing correction on the list of lists resulting from
enrichment analysis obtained using the runBatchGSE
and combineGSE functions.
This functions is based on the mt.rawp2adjp function
contained in the multtest package.
For each vector of p-value contained in the gseOut input object
a data.frame is returned, containing original p-value and corrected
p-values
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin, and Giovanni Parmigiani. "Integrating diverse genomic data using gene sets." Manuscript submitted.
###load gse analysis results for separate gene-to-phenotype score data(gseResultsSep) ###adjust for multiple testing using the Benjamini and Hochberg method gseABS.int.BH <- adjustPvalGSE(gseResultsSep) ###adjust for multiple testing using the Holm method gseABS.int.holm <- adjustPvalGSE(gseResultsSep, proc = "Holm")###load gse analysis results for separate gene-to-phenotype score data(gseResultsSep) ###adjust for multiple testing using the Benjamini and Hochberg method gseABS.int.BH <- adjustPvalGSE(gseResultsSep) ###adjust for multiple testing using the Holm method gseABS.int.holm <- adjustPvalGSE(gseResultsSep, proc = "Holm")
Combines GSE results obtained from a sepatate set of gene-to-phenotypes scores
combineGSE(gseOut, method)combineGSE(gseOut, method)
gseOut |
a list of lists containing the enrichment results
to be combined. This is usually the the output of
|
method |
character, this argument specifies the method
used to combine the enrichment results obtained from
distinct genes-to-phenotype scores (usually one per
genomic platform).
Available options are the computation of the geometric
or arithmetic means, the use of the median, the selection of
the minimun or the maximum enrichment score, and
the random selection of a score (respectively
|
This function summarize enrichment results obtained from distinct gene-to-phenotypes scores (usually one per genomic platform) by one of several alternative methods.
The output is a list of lists containing integrated enrichment results for all FGS collections
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin, and Giovanni Parmigiani. "Integrating diverse genomic data using gene sets." Manuscript submitted.
###load gse analysis results for separate gene-to-phenotype score data(gseResultsSep) ###combine enrichment score results using geometric mean gseABS.sep.geoMean <- combineGSE(gseResultsSep, method="geometricMean") ###combine enrichment score results using maximum value gseABS.sep.max <- combineGSE(gseResultsSep, method="max")###load gse analysis results for separate gene-to-phenotype score data(gseResultsSep) ###combine enrichment score results using geometric mean gseABS.sep.geoMean <- combineGSE(gseResultsSep, method="geometricMean") ###combine enrichment score results using maximum value gseABS.sep.max <- combineGSE(gseResultsSep, method="max")
computeDrStat Computes gene-to-phenotype associations scores,
using as input the output from convertToDr.
computeDrStat(data, columns = c(1:(ncol(data)-1)), method = "dev", integrate = TRUE)computeDrStat(data, columns = c(1:(ncol(data)-1)), method = "dev", integrate = TRUE)
data |
a list of data.frames containing genomic
measurements. Each element of |
columns |
a data.frame indicating patients' phenotypic class |
method |
character, the number of genomic platforms |
integrate |
logical, wheter to integrate the gene-to-phenotype scores across platform or return separates scores for each platform |
This function allows computing gene-to-phenotype association scores,
using as input the gene-centered list produced by computeDr.
The computeDrStat function works separately on each
gene-centered data.frame created by the convertToDr function,
assuming that the phenotype information is stored in the last column
named "response". It is possible computing both separate
association scores for each platform, as well as an integrated score,
as specified by the integrate arguments.
There are currently three methods available for obtaining the scores
(see Tyekucheva et al, manuscript under review),
as specified by the methods argument:
"dev": this approach computes the score as the difference of deviances;
"aic": this approach computes the score as the Akaike
information criterion for model selection;
"bic": this approach computes the score as the penalized likelihood ratio;
A list of named vectors containing separate or integrared gene-to-phenotype association scores.
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin, and Giovanni Parmigiani. "Integrating diverse genomic data using gene sets." Manuscript submitted.
###load data data(exampleData) ###convert dataDr <- convertToDr(dat, pheno, 4) ###compute the integrated score bicStatIntegrated <- computeDrStat(dataDr, columns = c(1:4), method="bic", integrate = TRUE) ###compute separate scores for each genomic platform bicStatSeparate <- computeDrStat(dataDr, columns = c(1:4), method="bic", integrate = FALSE)###load data data(exampleData) ###convert dataDr <- convertToDr(dat, pheno, 4) ###compute the integrated score bicStatIntegrated <- computeDrStat(dataDr, columns = c(1:4), method="bic", integrate = TRUE) ###compute separate scores for each genomic platform bicStatSeparate <- computeDrStat(dataDr, columns = c(1:4), method="bic", integrate = FALSE)
convertToDr converts genomic data into a list further
used for computing gene-to-phenotype association scores.
convertToDr(dataIntersection, response, nPlatforms = length(data))convertToDr(dataIntersection, response, nPlatforms = length(data))
dataIntersection |
a list of data.frames containing genomic
measurements. Each element of |
response |
a data.frame indicating patients' phenotypic class |
nPlatforms |
numeric, the number of genomic platforms |
This function coverts a list of data.frames containing distinct
genomic measurtements performed on the same patients
into a gene-centered used in further analyses for computing
gene-to-phenotype scores.
Data.frame in the input list (dataIntersection) must
have the same dimentions, with columns being patients,
and rows being genes. Column names identify the patients,
while rownames identify the genes.
The argument response is used to pass phenotypic information
about samples to be analyzed. This is a simple two columns
data.frame in which the first column correspond to patients
identifiers, and the second column to the phenotypic response
encoded as binary class (using the integers 0 and 1).
The nPlatforms argument specifies the number of platforms
that will be analyzed.
A list of data.frames, one for each analyzed gene, summarizing all genomic measurements and phenotipic information across patients and platforms.
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin,and Giovanni Parmigiani "Integrating diverse genomic data using gene sets." Manuscript submitted.
###load data data(exampleData) ###convert dataDr <- convertToDr(dat, pheno, 4)###load data data(exampleData) ###convert dataDr <- convertToDr(dat, pheno, 4)
A small subset of pre-processed Glioblasoma Multiforme (GBM)
genomic data from The Cancer Genome Atlas (TCGA) project,
encompassing Differential Gene Expression (DGE), and
Copy Number Variation (CNV).
Can be used as input to convertToDr.
data(exampleData)data(exampleData)
This object is a list of 4 data.frames
containing genomic measurements
obtained across distinct genomic scopes
(copy number variation and gene expression),
platforms (Affymetrix and Agiles), and laboratories.
In particular each data.frame consist of 500 gene measurements (by rows),
for 95 distinct patients (by columns) from the following 4 distinct platforms:
"dat.affy": DGE obtained using Affymetrix microarrays;
"dat.agilent": DGE obtained using Agilent microarrays;
"dat.cnvHarvard": CNV data obtained at Harvard;
"dat.cnvMskcc": CNV data obtained at Memorial Sloan Ketterng Cancer Center;
The Cancer Genome Atlas (TCGA) project http://cancergenome.nih.gov/
The Cancer Genome Atlas (TCGA) Research Network. “Comprehensive genomic characterization defines human glioblastoma genes and core pathways”. Nature, 2008, October 23; 455(7216): 1061-1068
data(exampleData) ls() class(dat) names(dat) sapply(dat,class) sapply(dat,dim)data(exampleData) ls() class(dat) names(dat) sapply(dat,class) sapply(dat,dim)
A list containing distinct types of FGS (i.e. Gene Ontology,
KEGG pathways). Each FGS is type is a list
of named chararacter vectors, one for each FGS,
containing the gene identifiers. Vectors names describe the FGS.
Can be used as input to runBatchGSE.
data(fgsList)data(fgsList)
This object is a list of length two:
"go": this is a list of 5 character vectors,
corresponding to 5 distinct Gene Ontology (GO) terms. Genes
annotated to each GO term are identified by their gene symbol;
"kegg": this is a list of 5 character vectors,
corresponding to 5 distinct KEGG pathways. Genes
annotated to each KEGG pathway are identified by their gene symbol;
The FGS were obtained from the org.Hs.eg.db pakage,
(use the org.Hs.eg function
to see the content);
These FGS were annotated using data from
GO.db and KEGGREST packages
(use the GO and
keggGet functions to see the content).
data(fgsList) class(fgsList) names(fgsList) str(fgsList)data(fgsList) class(fgsList) names(fgsList) str(fgsList)
A list containing distinct named numeric vectors
corresponding to the gene set enrichment p-value
separately computed with runBatchGSE
for each distinct data set.
Note that for each data set there are two set of p-values,
one for GO and one for KEGG.
These separate p-values can be combined across data sets
by the combineGSE function.
Can be used as input to adjustPvalGSE.
data(gseResultsSep)data(gseResultsSep)
This object is a list of length four:
"dat.affy": a list of legnth two:
"go" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 GO terms on Affymetrix gene expression data;
"keg" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 KEGG pathways on Affymetrix gene expression data;
"dat.agilent": a list of legnth two:
"go" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 GO terms on Agilent gene expression data;
"keg" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 KEGG pathways on Agilent gene expression data;
"dat.cnvHarvard": a list of legnth two:
"go" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 GO terms on Harvard CNV data;
"keg" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 KEGG pathways on Harvard CNV data;
"dat.cnvMskcc": a list of legnth two:
"go" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 GO terms on MSKCC CNV data;
"keg" is a numeric vector of length 5,
containing the p-values resulting from gene set enrichment
analysis of 5 KEGG pathways on MSKCC CNV data;
Computed using the runBatchGSE function
from the TCGA data contained in sepScores
and fgsList.
data(gseResultsSep) class(gseResultsSep) names(gseResultsSep) str(gseResultsSep)data(gseResultsSep) class(gseResultsSep) names(gseResultsSep) str(gseResultsSep)
A list containing a named numeric vector
corresponding to the genomic score resulting from the integration
across distinct data set.
These integrated gene-to-phenotype scores are computed
by the computeDrStat function.
Can be used as input to runBatchGSE.
data(intScores)data(intScores)
This object is a list of length one:
"integrated": a numeric vector of length 500,
corresponding to the integrated phenotype association scores
computed for each of the 500 genes used in the examples;
Computed using the computeDrStat function
from the TCGA data contained in dat and pheno.
data(intScores) class(intScores) names(intScores) str(intScores)data(intScores) class(intScores) names(intScores) str(intScores)
A data.frame with 2 columns containing the phenotypic class indicator
for the 95 patients analyzed and used in the examples.
Can be used as input to convertToDr.
data(exampleData)data(exampleData)
This object is a data.frame
with two columns:
"Sample": the first column contains the patients identifiers;
"Class": the second columns contain a numeric indicator
(0 or 1) corresponding to the phenotypic class of each patient;
The Cancer Genome Atlas (TCGA) project http://cancergenome.nih.gov/
The Cancer Genome Atlas (TCGA) Research Network. “Comprehensive genomic characterization defines human glioblastoma genes and core pathways”. Nature, 2008, October 23; 455(7216): 1061-1068
data(exampleData) class(pheno) colnames(pheno) str(pheno)data(exampleData) class(pheno) colnames(pheno) str(pheno)
The runBatchGSE function enables performing Gene Set
Enrichment analysis over multiple ranking statistics and multiple
lists of gene sets.
By default this function is an interface to the geneSetTest
in the limma package, and most of the arguments passed to
runBatchGSE are indeed passed to such lower
level function.
As an alternative the user can also define and pass to
runBatchGSE a custom function, defining the ranking statistics
and the gene set membership in the same way done
for geneSetTest (see Details below).
runBatchGSE(dataList, fgsList, ...)runBatchGSE(dataList, fgsList, ...)
dataList |
a list containing the gene-to-phenotype scores to be used
as ranking statistics in the GSE analysis. This list is usually
produced by running |
fgsList |
a list of FGS collection, in which each element is a list of character vectors, one for each gene set |
... |
additional arguments to be passed to lower level functions (see details below) |
This function performs enrichment analysis for all the gene-to-phenotype
scores (argument dataList) passed to it over a list of F
unctional Gene Set (FGS) (argument fgsList), returning
a p-value for each FGS.
Additional arguments can be bassed to this function to modify
the way the enrichment test is performed, as follows:
absolute logical, this specifies whether the absolute values of
the ranking statistics should be used in the test (the default
being TRUE)
gseFunc a function to perform GSE analysis. If not specified
the default is the geneSetTest function from the
limma package. If a function is specified by the user, the
membership of the analyzed genes to a FGS, and the ranking
statistics must be defined in the same way this is done for
geneSetTest, and the new function must
return an integer (usually a p-value) (see the help for
geneSetTest)
The following main arguments are used by geneSetTest:
type character, specifies the type of statistics used to rank
the genes by geneSetTest: 'f' for F-like statistics
(default), 't' for t-like statistics, or 'auto' for an
educated guess
alternative character, defines the alternative with the
following possible options: 'mixed' (default),
'either', 'up' or 'down',
'two.sided', 'greater', or 'less'
ranks.only logical, if TRUE (default) only ranks will be
used by geneSetTest
nsim numeric, the number of randomly selected sets of genes to
be used in simulations to compute the p-value
The output is a list of lists containing the set of enrichment results for all gene-to-phenotype scores and FGS collections used as input.
Luigi Marchionni [email protected]
Svitlana Tyekucheva, Luigi Marchionni, Rachel Karchin, and Giovanni Parmigiani. "Integrating diverse genomic data using gene sets." Manuscript submitted.
###require limma to run the example require(limma) ###load integrated gene-to-phenotype scores data(intScores) ###load separate gene-to-phenotype scores data(sepScores) ###load list of functional gene sets data(fgsList) ###run GSE analysis in batch with default parameters gseABS.int <- runBatchGSE(dataList=intScores, fgsList=fgsList) ###run GSE analysis in batch with alternative parameters gseABS.sep <- runBatchGSE(dataList=sepScores, fgsList=fgsList, absolute=FALSE, type="t", alternative="up") ###run GSE analysis in batch passing an enrichment function gseUP.int.2 <- runBatchGSE(dataList=intScores, fgsList=fgsList, absolute=FALSE, gseFunc=wilcoxGST, alternative="up") ###define and use a new enrichment function gseFunc <- function (selected, statistics, threshold) { diffExpGenes <- statistics > threshold tab <- table(diffExpGenes, selected) pVal <- fisher.test(tab)[["p.value"]] } gseUP.sep.2 <- runBatchGSE(dataList=sepScores, fgsList=fgsList, absolute=FALSE, gseFunc=gseFunc, threshold=7.5)###require limma to run the example require(limma) ###load integrated gene-to-phenotype scores data(intScores) ###load separate gene-to-phenotype scores data(sepScores) ###load list of functional gene sets data(fgsList) ###run GSE analysis in batch with default parameters gseABS.int <- runBatchGSE(dataList=intScores, fgsList=fgsList) ###run GSE analysis in batch with alternative parameters gseABS.sep <- runBatchGSE(dataList=sepScores, fgsList=fgsList, absolute=FALSE, type="t", alternative="up") ###run GSE analysis in batch passing an enrichment function gseUP.int.2 <- runBatchGSE(dataList=intScores, fgsList=fgsList, absolute=FALSE, gseFunc=wilcoxGST, alternative="up") ###define and use a new enrichment function gseFunc <- function (selected, statistics, threshold) { diffExpGenes <- statistics > threshold tab <- table(diffExpGenes, selected) pVal <- fisher.test(tab)[["p.value"]] } gseUP.sep.2 <- runBatchGSE(dataList=sepScores, fgsList=fgsList, absolute=FALSE, gseFunc=gseFunc, threshold=7.5)
A list containing distinct named numeric vectors
corresponding to the gene-to-phenotype association scores
resulting from the sepatate analysis of each data set.
These separate gene-to-phenotype scores are compurted
by computeDrStat function.
Can be used as input to runBatchGSE.
data(sepScores)data(sepScores)
This object is a list of length four, one element for data set:
"dat.affy": a numeric vector of length 500,
corresponding to the sepatate phenotype association scores
computed for Affymetrix gene expression data;
"dat.agilent": a numeric vector of length 500,
corresponding to the separate phenotype association scores
computed for Agilent gene expression data;
"dat.cnvHarvard": a numeric vector of length 500,
corresponding to the separate phenotype association scores
computed for Harvard CNV data;
"dat.cnvMskcc": a numeric vector of length 500,
corresponding to the separate phenotype association scores
computed for the MSKCC CNV data;
Computed using the computeDrStat function
from the TCGA data contained in data.
data(sepScores) class(sepScores) names(sepScores) str(sepScores)data(sepScores) class(sepScores) names(sepScores) str(sepScores)