| Title: | PRojection Onto the Most Interesting Statistical Evidence |
|---|---|
| Description: | A general tool to identify genomic features with a specific biologically interesting pattern of associations with multiple endpoint variables as described in Pounds et. al. (2009) Bioinformatics 25: 2013-2019 |
| Authors: | Stan Pounds <[email protected]>, Xueyuan Cao <[email protected]> |
| Maintainer: | Stan Pounds <[email protected]>, Xueyuan Cao <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 1.59.0 |
| Built: | 2024-11-30 03:18:08 UTC |
| Source: | https://github.com/bioc/PROMISE |
a tool to identify genomic geatures with a specific biologically interesting pattern of associations with multiple endpoint variables
| Package: | PROMISE |
| Type: | Package |
| Version: | 1.17.0 |
| Date: | 2014-6-24 |
| License: | GPL (>=2) |
| LazyLoad: | yes |
The PROMISE (PRojection Onto the Most Interesting Statistical Evidence) is performed by calling function PROMISE. The array data and endpoint data are passed through an ExpressionSet; the gene set definition is passed through a GeneSetCollection, and PROMISE definition is passed through a data frame. promise.genestat and avg.abs.genestat are called internally by PROMISE. Two R routines for calculating association statistics with individual endpoint variable(jung.rstat and spearman.rstat) are provided in this version. Users could provide their own R routines written in a similar fashion.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Maintainer: Stan Pound [email protected]; Xueyuan Cao [email protected]
Jung, S-H, Owzar K, and Goerge SL (2005) A multiple testing procedure to associate gene expression levels with survival. Biostatistics 24: 3077-3088.
Goeman JJ and Buhlmann P (2007) Analyzing gene expression data in terms of gene sets: methodological issues. Bioinformatics 23: 980-987.
Pounds S, Cheng C, Cao X, Crews KR, Plunkett W, Gandhi V, Rubnitz J, Ribeiro RC, Downing JR, and Lamba J (2009) PROMISE: a tool to identify genomic features with a specific biologically interesting pattern of associations with multiple endpoint variables. Bioinformatics 25: 2013-2019
## load sampExprSet, sampGeneSet, phPatt. data(sampExprSet) data(sampGeneSet) data(phPatt) ## Perform PROMISE procedure without GSEA test1 <- PROMISE(exprSet=sampExprSet, geneSet=NULL, promise.pattern=phPatt, strat.var=NULL, proj0=FALSE, nbperm=FALSE, max.ntail=10, seed=13, nperms=100) ## Perform PROMISE procedure with GSEA and using fast permuation test2 <- PROMISE(exprSet=sampExprSet, geneSet=sampGeneSet, promise.pattern=phPatt, strat.var=NULL, proj0=TRUE, nbperm=TRUE, max.ntail=10, seed=13, nperms=100)## load sampExprSet, sampGeneSet, phPatt. data(sampExprSet) data(sampGeneSet) data(phPatt) ## Perform PROMISE procedure without GSEA test1 <- PROMISE(exprSet=sampExprSet, geneSet=NULL, promise.pattern=phPatt, strat.var=NULL, proj0=FALSE, nbperm=FALSE, max.ntail=10, seed=13, nperms=100) ## Perform PROMISE procedure with GSEA and using fast permuation test2 <- PROMISE(exprSet=sampExprSet, geneSet=sampGeneSet, promise.pattern=phPatt, strat.var=NULL, proj0=TRUE, nbperm=TRUE, max.ntail=10, seed=13, nperms=100)
A function to calculate the mean of absolute values of statistics based on a gene set definition
avg.abs.genestat(gene.res, probes, GS.data)avg.abs.genestat(gene.res, probes, GS.data)
gene.res |
a data frame. Each row gives test statistics for a genomic variable. Each column corresponds to an endpoint variable. |
probes |
a vector that links the gene.res to GS.data. |
GS.data |
a data frame with first column for probe set identifier and second column for gene set identifier. Each row assigns a probe set to a gene set. Each probe set may be assigned to multiple gene sets or no gene set at all. |
Return a matrix of statistics. Each row gives the mean absolute value of test statistics of genes belonging to a gene set. The columns are same as in gene.res.
A function internally called by PROMISE.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Goeman JJ and Buhlmann P (2007) Analyzing gene expression data in terms of gene sets: methodological issues. Bioinformatics 23: 980-987.
## load sampExprSet sampGeneSet. data(sampExprSet) data(sampGeneSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) probes <- rownames(Y) ## convert sampGeneSet to a data frame GS.data <- NULL for (i in 1:length(sampGeneSet)){ tt <- sampGeneSet[i][[1]] this.name <- unlist(geneIds(tt)) this.set <- setName(tt) GS.data <- rbind.data.frame(GS.data, cbind.data.frame(featureID=as.character(this.name), setID=rep(as.character(this.set), length(this.name)))) } ## Calculate the mean of absolute values of statistics ## This is only a demo, probe expression values are used ##in stead of statistics test <- avg.abs.genestat(Y, probes, GS.data)## load sampExprSet sampGeneSet. data(sampExprSet) data(sampGeneSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) probes <- rownames(Y) ## convert sampGeneSet to a data frame GS.data <- NULL for (i in 1:length(sampGeneSet)){ tt <- sampGeneSet[i][[1]] this.name <- unlist(geneIds(tt)) this.set <- setName(tt) GS.data <- rbind.data.frame(GS.data, cbind.data.frame(featureID=as.character(this.name), setID=rep(as.character(this.set), length(this.name)))) } ## Calculate the mean of absolute values of statistics ## This is only a demo, probe expression values are used ##in stead of statistics test <- avg.abs.genestat(Y, probes, GS.data)
Compute statistic that measures the correlation of many continuous variables with a censored time-to-event variable
jung.rstat(x, time.cens, strat = NULL)jung.rstat(x, time.cens, strat = NULL)
x |
a data frame with row corresponding to probe set and column corresponding to subjects, the order of columns (subjects) should match the order of rows in time.cens. |
time.cens |
a data frame with number of row equal to number of column in x. It contains two columns with first for time and second for censor (1 = event, 0 = censored). |
strat |
a vector of stratum to calculate stratified r-type association statistics, default = NULL. |
Returns a vector of Jun's r-type association statistics.
The order of subjects in x (column), time.cens, and strat should all match. The original statistic proposed by Jung, Owzar, and George can be written as a dot-product. The statistic returned by this routine is expressed in the form of a correlation statistic by dividing the dot product by the square root of the lengths of the two vectors in the numerator.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Jung SH, Owzar K, and George SL (2005) A multiple testing procedure to associate gene expression levels with survival. Stat Med 24:3077-88
## load sampExprSet. data(sampExprSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) ## extract end point data from sampExprSet time.cens <-pData(phenoData(sampExprSet))[, 3:4] strat <- pData(phenoData(sampExprSet))$strat ## compute Jung's r-type association statistics jungstat <- jung.rstat(Y, time.cens, strat = strat)## load sampExprSet. data(sampExprSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) ## extract end point data from sampExprSet time.cens <-pData(phenoData(sampExprSet))[, 3:4] strat <- pData(phenoData(sampExprSet))$strat ## compute Jung's r-type association statistics jungstat <- jung.rstat(Y, time.cens, strat = strat)
This hypothetical phenotype pattern definition set phPatt has three columns: stat.coef, stat.func, and endpt.vars. It defines an associatin pattern for three phenotypes.
data(phPatt)data(phPatt)
Perform permutation-based test to identify genes with expression levels having a specific biologically interesting pattern of associations with multiple endpoint variables
PROMISE(exprSet, geneSet=NULL, promise.pattern, strat.var=NULL, proj0=FALSE, seed=13, nbperm=FALSE, max.ntail=100, nperms=10000)PROMISE(exprSet, geneSet=NULL, promise.pattern, strat.var=NULL, proj0=FALSE, seed=13, nbperm=FALSE, max.ntail=100, nperms=10000)
exprSet |
an ExpressionSet class contains minimum of exprs (expression matrix) and phenoData (AnnotatedDataFrame of end point data). Please refer to Biobase for details on how to create such an ExpressionSet. |
geneSet |
a GeneSetCollection class with minimum of setName and geneIDs for each GeneSet. Please refer to GSEABase for how to create such a GeneSetCollection class. The default is NULL which will perform no gene set enrichment analysis. |
promise.pattern |
a data frame defining the association pattern of interest. The column names must be stat.coef, stat.func, and endpt.vars. The stat.coef column gives the coefficients for combining the statistics of association of genomic variables with individual endpoint variable into the ultimate PROMISE statistic. If proj0=TRUE, the stat.coeff is ignored. The stat.func column gives the name of the R routine that computes the test statistic of association of the endpoint variables. Two R routines (jung.rstat and spearman.rstat)are provided. Users can provide their own routine accordingly. The endpt.vars column gives the name(s) of variable(s) in the endpoint data file needed to compute each term of the PROMISE statistic. A common without a space should be used to separate multiple variables that correspond to the same term in the association pattern definition. |
strat.var |
the name or numeric value of stratum variable in exprSet for stratified analysis. The default is NULL which performs an unstratified analysis. |
proj0 |
indicator of whether projection to 0 is performed. It takes two valid values: TRUE or FALSE. If proj0=TRUE, PROMISE statistics is the sum of squares of individual statitics and the stat.coeff in promise.pattern is ignored. The default is FALSE. |
seed |
initial seed of random number generator. The default is 13. |
nbperm |
indicator of fast permuation using negative binomial strategy, taking two valid values: FALSE or TRUE. The default is FALSE. |
max.ntail |
number of sucess if nbperm = T. Further permutation will not be performed for gene(s) or gene set(s) which max.ntail permutated statistics are greater or equal to the observed statistics, The default is 100 |
nperms |
number of permutations. The default is 10,000. |
$generes |
individual genes' test statistics and p-values for each individual endpoint and PROMISE analysis. If nbperm=T, the last column contains number of permuations for each gene. |
$setres |
gene set's test statistics and p-values for each individual endpoint and PROMISE analysis. If nbperm=T, the last column contains number of permuations for gene set. If geneSet is NULL, the value of this component is also NULL. |
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Pounds S, Cheng C, Cao X, Crews KR, Plunkett W, Gandhi V, Rubnitz J, Ribeiro RC, Downing JR, and Lamba J (2009) PROMISE: a tool to identify genomic features with a specific biologically interesting pattern of associations with multiple endpoint variables. Bioinformatics 25: 2013-2019
jung.rstat avg.abs.genestat
promise.genestat spearman.rstat
promise.pattern
## load sampExprSet, sampGeneSet, phPatt. data(sampExprSet) data(sampGeneSet) data(phPatt) ## Perform PROMISE procedure without GSEA test1 <- PROMISE(exprSet=sampExprSet, geneSet=NULL, promise.pattern=phPatt, strat.var=NULL, proj0=FALSE, nbperm=FALSE, max.ntail=10, seed=13, nperms=100) ## Perform PROMISE procedure with GSEA and using fast permuation test2 <- PROMISE(exprSet=sampExprSet, geneSet=sampGeneSet, promise.pattern=phPatt, strat.var=NULL, proj0=TRUE, nbperm=TRUE, max.ntail=10, seed=13, nperms=100)## load sampExprSet, sampGeneSet, phPatt. data(sampExprSet) data(sampGeneSet) data(phPatt) ## Perform PROMISE procedure without GSEA test1 <- PROMISE(exprSet=sampExprSet, geneSet=NULL, promise.pattern=phPatt, strat.var=NULL, proj0=FALSE, nbperm=FALSE, max.ntail=10, seed=13, nperms=100) ## Perform PROMISE procedure with GSEA and using fast permuation test2 <- PROMISE(exprSet=sampExprSet, geneSet=sampGeneSet, promise.pattern=phPatt, strat.var=NULL, proj0=TRUE, nbperm=TRUE, max.ntail=10, seed=13, nperms=100)
a function to calculate individual gene and PROMISE statistics for a defined pattern of association
promise.genestat(Y, ph.data, ph.pattern, strat = NULL, proj0=FALSE)promise.genestat(Y, ph.data, ph.pattern, strat = NULL, proj0=FALSE)
Y |
a data frame with row corresponding to probe set and column corresponding to subjects, the order of column should match order of row in ph.data. |
ph.data |
a data frame with rows corresponding to subjects and columns corresponding to endpoint variables. |
ph.pattern |
a data frame with column headers: stat.coef, stat.func, endpt.vars. The stat.coeff column gives the coefficients for combining the statistics of association of genomic variable with individual endpoint variable into the ultimate PROMISE statistic. If proj0=TRUE, the stat.coeff is ignored. The stat.func column gives the name of the R routine that computes the test statistic of association of the end point variables. jung.rstat and spearman.rstat are provided. Users can provide their own routines accordingly. The endpt.vars column gives the name(s) of variable(s) in ph.data needed to compute each term of the PROMISE statistic. A comma without a space should be used to separate multiple variables that correspond to the same term in the association pattern definition. |
strat |
a vector of stratum to calculate stratified statistics. The default is NULL. |
proj0 |
indicator of whether projection to 0 is performed. It takes two valid values: TRUE or FALSE. If proj0=TRUE, PROMISE statistics is the sum of squares of individual statitics. The default is FALSE. |
a matrix of statistics. Each row gives gene's statistics of each individual endpoint and the PROMISE statistics defined in ph.pattern.
a function internally called by PROMISE.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Pounds S, Cheng C, Cao X, Crews KR, Plunkett W, Gandhi V, Rubnitz J, Ribeiro RC, Downing JR, and Lamba J (2009) PROMISE: a tool to identify genomic features with a specific biologically interesting pattern of associations with multiple endpoint variables. Bioinformatics 25: 2013-2019
## load sampExprSet, phPatt. data(sampExprSet) data(phPatt) Y <- exprs(sampExprSet) ph.data <- pData(phenoData(sampExprSet)) test <- promise.genestat(Y, ph.data, phPatt, strat=ph.data[, 5]) test2 <- promise.genestat(Y, ph.data, phPatt, strat=ph.data[, 5], proj0=TRUE)## load sampExprSet, phPatt. data(sampExprSet) data(phPatt) Y <- exprs(sampExprSet) ph.data <- pData(phenoData(sampExprSet)) test <- promise.genestat(Y, ph.data, phPatt, strat=ph.data[, 5]) test2 <- promise.genestat(Y, ph.data, phPatt, strat=ph.data[, 5], proj0=TRUE)
PROMISE pattern is a data frame of association pattern definition, consisting of three columns.
PROMISE pattern: The column names must be stat.coef, stat.func, and endpt.vars.
stat.coef column gives the coefficients for combining the statistics of association of genomic variable with individual endpoint variable into the ultimate PROMISE statistic.
stat.func column gives the name of the R routine that computes the test statistic of association of the endpoint variables. Two R routines (jung.rstat and spearman.rstat)are provided in current release. Users can provide their own routine accordingly.
endpt.vars column gives the name(s) of variable(s) in the endpoint data frame needed to compute each term of the PROMISE statistic. If more than one variables involve in one term, they should be separated by a comma without space.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Pounds S, Cheng C, Cao X, Crews KR, Plunkett W, Gandhi V, Rubnitz J, Ribeiro RC, Downing JR, and Lamba J (2009) PROMISE: a tool to identify genomic features with a specific biologically interesting pattern of associations with multiple endpoint variables. Bioinformatics 25: 2013-2019
This hypothetical expression set sampExpSet belongs to an ExpressionSet class. It contains 100 genomic features (probe_1 to probe_100) for 50 subjects (array_1 to array_50) and phenotype data of drugLevel, residualDisease, obsTime, obsCensor and strat. The expression values can be accessed by exprs(sampExprSet). The phenotype data can be accessed by pData(phenoData(sampExprSet))
data(sampExprSet)data(sampExprSet)
This hypothetical gene set sampGeneSet belongs to a GeneSetCollection class. It contains 10 gene sets (GeneSet class).
data(sampGeneSet)data(sampGeneSet)
A function to calculate Spearman rank correlation of each gene in an array data with a continuous variable
spearman.rstat(Y, x, strat = NULL)spearman.rstat(Y, x, strat = NULL)
Y |
a numeric data frame. Each row gives values of one genomic variable. |
x |
a vector of continuous variable. |
strat |
a vector of stratum to calculate stratified correlation statistics, default = NULL. |
Return a vector of Spearman rank correlation statistics.
Stan Pounds [email protected]; Xueyuan Cao [email protected]
Spearman C. (1904) The proof and measurement of association between two things. Amer. J. Psychol. 15: 72-101
## load sampExprSet. data(sampExprSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) ## extract end point data from sampExprSet x <- pData(phenoData(sampExprSet))$drugLevel strat <- pData(phenoData(sampExprSet))$strat ## Calculte Spearman correlation statistics test <- spearman.rstat(Y, x, strat = strat)## load sampExprSet. data(sampExprSet) ## extract expression matrix from sampExprSet Y <- exprs(sampExprSet) ## extract end point data from sampExprSet x <- pData(phenoData(sampExprSet))$drugLevel strat <- pData(phenoData(sampExprSet))$strat ## Calculte Spearman correlation statistics test <- spearman.rstat(Y, x, strat = strat)