Description Usage Arguments Details Value Methods (by class) Note Author(s) References See Also Examples
wmwTest is a highly efficient Wilcoxon-Mann-Whitney rank sum
test for high-dimensional data, such as gene expression profiling. For datasets with
more than 100 features (genes), the function can be more than 1,000
times faster than its R implementations (wilcox.test
in
stats
, or rankSumTestWithCorrelation
in limma
).
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 | wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = c("p.greater", "p.less", "p.two.sided", "U", "abs.log10p.greater",
"log10p.less", "abs.log10p.two.sided", "Q", "r", "f", "U1", "U2"),
simplify = TRUE
)
## S4 method for signature 'matrix,IndexList'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'numeric,IndexList'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'matrix,GmtList'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'eSet,GmtList'
wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = "p.greater",
simplify = TRUE
)
## S4 method for signature 'eSet,numeric'
wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = "p.greater",
simplify = TRUE
)
## S4 method for signature 'eSet,logical'
wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = "p.greater",
simplify = TRUE
)
## S4 method for signature 'eSet,list'
wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = "p.greater",
simplify = TRUE
)
## S4 method for signature 'ANY,numeric'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'ANY,logical'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'ANY,list'
wmwTest(x, indexList, valType = "p.greater", simplify = TRUE)
## S4 method for signature 'matrix,SignedIndexList'
wmwTest(x, indexList, valType, simplify = TRUE)
## S4 method for signature 'matrix,SignedGenesets'
wmwTest(x, indexList, valType, simplify = TRUE)
## S4 method for signature 'numeric,SignedIndexList'
wmwTest(x, indexList, valType, simplify = TRUE)
## S4 method for signature 'eSet,SignedIndexList'
wmwTest(x, indexList, valType, simplify = TRUE)
## S4 method for signature 'eSet,SignedGenesets'
wmwTest(
x,
indexList,
col = "GeneSymbol",
valType = c("p.greater", "p.less", "p.two.sided", "U", "abs.log10p.greater",
"log10p.less", "abs.log10p.two.sided", "Q", "r", "f", "U1", "U2"),
simplify = TRUE
)
|
x |
A numeric matrix. All other data types (e.g. numeric vectors
or |
indexList |
A list of integer indices (starting from 1) indicating
signature genes. Can be of length zero. Other data types (e.g. a list
of numeric or logical vectors, or a numeric or logical vector) are
coerced into such a list. See |
col |
a string sometimes used with a |
valType |
The value type to be returned, allowed values
include |
simplify |
Logical. If not, the returning value is in matrix
format; if set to |
The basic application of the function is to test the enrichment of gene sets in expression profiling data or differentially expressed data (the matrix with feature/gene in rows and samples in columns).
A special case is when x
is an eSet
object
(e.g. ExpressionSet
), and indexList
is a list returned
from readGmt
function. In this case, the only requirement is
that one column named GeneSymbol
in the featureData
contain gene symbols used in the GMT file. The same applies to signed Gmt files. See the example below.
Besides the conventional value types such as ‘p.greater’,
‘p.less’, ‘p.two.sided’ , and ‘U’ (the U-statistic),
wmwTest
(from version 0.99-1) provides further value types:
abs.log10p.greater
and log10p.less
perform log10
transformation on respective p-values and give the
transformed value a proper sign (positive for greater than, and
negative for less than); abs.log10p.two.sided
transforms
two-sided p-values to non-negative values; and Q
score
reports absolute log10-transformation of p-value of the
two-side variant, and gives a proper sign to it, depending on whether it is
rather greater than (positive) or less than (negative).
From version 1.19.1, the rank-biserial correlation coefficient (‘r’) and the common language effect size (‘f’) are supported value types.
Before version 1.19.3, the ‘U’ statistic returned is in fact ‘U2’. From version 1.19.3, ‘U1’ is returned when ‘U’ is used, and users can specify additional parameter values ‘U1’ and ‘U2’. The sum of ‘U1’ and ‘U2’ is the product of the sizes of two vectors to be compared.
A numeric matrix or vector containing the statistic.
x = matrix,indexList = IndexList
: x
is a matrix
and indexList
is a IndexList
x = numeric,indexList = IndexList
: x
is a numeric
and indexList
is a IndexList
x = matrix,indexList = GmtList
: x
is a matrix
and indexList
is a GmtList
x = eSet,indexList = GmtList
: x
is a eSet
and indexList
is a GmtList
x = eSet,indexList = numeric
: x
is a eSet
and indexList
is a numeric
x = eSet,indexList = logical
: x
is a eSet
and indexList
is a logical
x = eSet,indexList = list
: x
is a eSet
and indexList
is a list
x = ANY,indexList = numeric
: x
is ANY
and indexList
is a numeric
x = ANY,indexList = logical
: x
is ANY
and indexList
is a logical
x = ANY,indexList = list
: x
is ANY
and indexList
is a list
x = matrix,indexList = SignedIndexList
: x
is a matrix
and indexList
is a
SignedIndexList
x = matrix,indexList = SignedGenesets
: x
is a eSet
and indexList
is a
SignedIndexList
x = numeric,indexList = SignedIndexList
: x
is a numeric
and indexList
is a
SignedIndexList
x = eSet,indexList = SignedIndexList
: x
is a eSet
and indexList
is a
SignedIndexList
x = eSet,indexList = SignedGenesets
: x
is a eSet
and indexList
is a
SignedIndexList
The function has been optimized for expression profiling data. It
avoids repetitive ranking of data as done by native R implementations
and uses efficient C code to increase the performance and control
memory use. Simulation studies using expression profiles of 22000
genes in 2000 samples and 200 gene sets suggested that the C
implementation can be >1000 times faster than the R
implementation. And it is possible to further accelerate by
parallel calling the function with mclapply
in the multicore
package.
Jitao David Zhang <jitao_david.zhang@roche.com>, with critical inputs from Jan Aettig and Iakov Davydov about U statistics.
Barry, W.T., Nobel, A.B., and Wright, F.A. (2008). A statistical framework for testing functional namespaces in microarray data. _Annals of Applied Statistics_ 2, 286-315.
Wu, D, and Smyth, GK (2012). Camera: a competitive gene set test accounting for inter-gene correlation. _Nucleic Acids Research_ 40(17):e133
Zar, JH (1999). _Biostatistical Analysis 4th Edition_. Prentice-Hall International, Upper Saddle River, New Jersey.
codewilcox.test in the stats
package, and rankSumTestWithCorrelation
in
the limma
package.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | ## R-native data structures
set.seed(1887)
rd <- rnorm(1000)
rl <- sample(c(TRUE, FALSE), 1000, replace=TRUE)
wmwTest(rd, rl, valType="p.two.sided")
wmwTest(rd, which(rl), valType="p.two.sided")
rd1 <- rd + ifelse(rl, 0.5, 0)
wmwTest(rd1, rl, valType="p.greater")
wmwTest(rd1, rl, valType="U")
rd2 <- rd - ifelse(rl, 0.2, 0)
wmwTest(rd2, rl, valType="p.greater")
wmwTest(rd2, rl, valType="p.two.sided")
wmwTest(rd2, rl, valType="p.less")
wmwTest(rd2, rl, valType="r")
wmwTest(rd2, rl, valType="f")
## matrix forms
rmat <- matrix(c(rd, rd1, rd2), ncol=3, byrow=FALSE)
wmwTest(rmat, rl, valType="p.two.sided")
wmwTest(rmat, rl, valType="p.greater")
wmwTest(rmat, which(rl), valType="p.two.sided")
wmwTest(rmat, which(rl), valType="p.greater")
## other valTypes
wmwTest(rmat, which(rl), valType="U")
wmwTest(rmat, which(rl), valType="abs.log10p.greater")
wmwTest(rmat, which(rl), valType="log10p.less")
wmwTest(rmat, which(rl), valType="abs.log10p.two.sided")
wmwTest(rmat, which(rl), valType="Q")
wmwTest(rmat, which(rl), valType="r")
wmwTest(rmat, which(rl), valType="f")
## using ExpressionSet
data(sample.ExpressionSet)
testSet <- sample.ExpressionSet
fData(testSet)$GeneSymbol <- paste("GENE_",1:nrow(testSet), sep="")
mySig1 <- sample(c(TRUE, FALSE), nrow(testSet), prob=c(0.25, 0.75), replace=TRUE)
wmwTest(testSet, which(mySig1), valType="p.greater")
## using integer
exprs(testSet)[,1L] <- exprs(testSet)[,1L] + ifelse(mySig1, 50, 0)
wmwTest(testSet, which(mySig1), valType="p.greater")
## using lists
mySig2 <- sample(c(TRUE, FALSE), nrow(testSet), prob=c(0.6, 0.4), replace=TRUE)
wmwTest(testSet, list(first=mySig1, second=mySig2))
## using GMT file
gmt_file <- system.file("extdata/exp.tissuemark.affy.roche.symbols.gmt", package="BioQC")
gmt_list <- readGmt(gmt_file)
gss <- sample(unlist(sapply(gmt_list, function(x) x$genes)), 1000)
eset<-new("ExpressionSet",
exprs=matrix(rnorm(10000), nrow=1000L),
phenoData=new("AnnotatedDataFrame", data.frame(Sample=LETTERS[1:10])),
featureData=new("AnnotatedDataFrame",data.frame(GeneSymbol=gss)))
esetWmwRes <- wmwTest(eset ,gmt_list, valType="p.greater")
summary(esetWmwRes)
## using signed GMT file
signed_gmt_file <- system.file("extdata/test.gmt", package="BioQC")
signed_gmt <- readSignedGmt(signed_gmt_file)
esetSignedWmwRes <- wmwTest(eset, signed_gmt, valType="p.greater")
esetMat <- exprs(eset); rownames(esetMat) <- fData(eset)$GeneSymbol
esetSignedWmwRes2 <- wmwTest(esetMat, signed_gmt, valType="p.greater")
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.