R/dot-permTestDiffAsso.R
In stefpeschel/NetCoMi: Network Construction and Comparison for Microbiome Data
Defines functions
.diffConnectNetwork
.diffConnectVariables
.diffConnectPairs
.permTestDiffAsso
Documented in
.permTestDiffAsso
#' @title Permutation Tests for Determining Differential Associations
#'
#' @description The function implements procedures to test whether pairs of taxa
#' are differentially associated, whether a taxon is differentially associated
#' to all other taxa, or whether two networks are differentially associated
#' between two groups as proposed by \cite{Gill et al.(2010)}.
#'
#' @param countMat1,countMat2 matrices containing microbiome data (read counts)
#' of group 1 and group 2 (rows represent samples and columns taxonomic units,
#' respectively).
#' @param countsJoint joint count matrices before preprocessing
#' @param normCounts1,normCounts2 normalized count matrices.
#' @param assoMat1,assoMat2 association matrices corresponding to the two count
#' matrices. The associations must have been estimated from the count matrices
#' \code{countMat1} and \code{countMat2}.
#' @param paramsNetConstruct parameters used for network construction.
#' @param method character vector indicating the tests to be performed. Possible
#' values are \code{"connect.pairs"} (differentially correlated taxa pairs),
#' \code{"connect.variables"} (one taxon to all other) and
#' \code{"connect.network"} (differentially connected networks). By default,
#' all three tests are conducted.
#' @param fisherTrans logical indicating whether the correlation values should
#' be Fisher-transformed.
#' @param pvalsMethod currently only \code{"pseudo"} is available, where 1 is
#' added to the number of permutations and the permutation test statistics
#' being more extreme than the observed one in order to avoid zero p-values.
#' @param adjust multiple testing adjustment for the tests for differentially
#' correlated pairs of taxa; possible values are "lfdr" (default) for local
#' false discovery rate correction (via \code{\link[fdrtool]{fdrtool}}) or one
#' of the methods provided by \code{\link[stats]{p.adjust}}
#' @param adjust2 multiple testing adjustment for the tests if a taxa pair is
#' differentially correlated to all other taxa; possible methods are those
#' provided by \code{\link[stats]{p.adjust}} (a few hundred tests are
#' necessary for the local fdr correction)
#' @param trueNullMethod character indicating the method used for estimating the
#' proportion of true null hypotheses from a vector of p-values. Used for the
#' adaptive Benjamini-Hochberg method for multiple testing adjustment (chosen
#' by \code{adjust = "adaptBH"}).
#' @param alpha significance level
#' @param lfdrThresh defines a threshold for the local fdr if "lfdr" is chosen
#' as method for multiple testing correction; defaults to 0.2, which means
#' that correlations with a corresponding local fdr less than or equal to 0.2
#' are identified as significant
#' @param nPerm number of permutations
#' @param matchDesign Numeric vector with two elements specifying an optional
#' matched-group (i.e. matched-pair) design, which is used for the permutation
#' tests in \code{\link{netCompare}} and \code{\link{diffnet}}. \code{c(1,1)}
#' corresponds to a matched-pair design. A 1:2 matching, for instance, is
#' defined by \code{c(1,2)}, which means that the first sample of group 1 is
#' matched to the first two samples of group 2 and so on.
#' The appropriate order of samples must be ensured. If
#' \code{NULL}, the group memberships are shuffled randomly while group sizes
#' identical to the original data set are ensured.
#' @param callNetConstr call inherited from \code{netConstruct()}.
#' @param cores number of CPU cores (permutation tests are executed parallel)
#' @param verbose if \code{TRUE}, status messages and numbers of SparCC
#' iterations are printed
#' @param logFile character string naming the log file within which the current
#' iteration number is stored
#' @param seed an optional seed for reproducibility of the results
#' @param fileLoadAssoPerm character giving the name (without extenstion)
#' or path of the file storing the "permuted" association/dissimilarity
#' matrices that have been exported by setting \code{storeAssoPerm} to
#' \code{TRUE}. Only used for permutation tests. Set to \code{NULL} if no
#' existing associations should be used.
#' @param fileLoadCountsPerm character giving the name (without extenstion)
#' or path of the file storing the "permuted" count matrices that have been
#' exported by setting \code{storeCountsPerm} to \code{TRUE}.
#' Only used for permutation tests, and if \code{fileLoadAssoPerm = NULL}.
#' Set to \code{NULL} if no existing count matrices should be used.
#' @param storeAssoPerm logical indicating whether the association (or
#' dissimilarity) matrices for the permuted data should be stored in a file.
#' The filename is given via \code{fileStoreAssoPerm}. If \code{TRUE},
#' the computed "permutation" association/dissimilarity matrices can be reused
#' via \code{fileLoadAssoPerm} to save runtime. Defaults to \code{FALSE}.
#' @param fileStoreAssoPerm character giving the file name to store a matrix
#' containing a matrix with associations/dissimilarities for the permuted
#' data. Can also be a path.
#' @param storeCountsPerm logical indicating whether the permuted count matrices
#' should be stored in an external file. Defaults to \code{FALSE}.
#' @param fileStoreCountsPerm character vector with two elements giving the
#' names of two files storing the permuted count matrices belonging to the
#' two groups.
#' @param assoPerm not used anymore.
#' @references
#' \insertRef{gill2010statistical}{NetCoMi}\cr\cr
#' \insertRef{knijnenburg2009fewer}{NetCoMi}\cr\cr
#'
#' @seealso \code{\link{diffnet}}
.permTestDiffAsso <- function(countMat1, countMat2, countsJoint,
normCounts1, normCounts2,
assoMat1, assoMat2,
paramsNetConstruct,
method = c("connect.pairs",
"connect.variables",
"connect.network"),
fisherTrans = TRUE,
pvalsMethod = "pseudo",
adjust = "lfdr",
adjust2 = "holm",
trueNullMethod = "convest",
alpha = 0.05,
lfdrThresh = 0.2,
nPerm = 1000,
matchDesign = NULL,
callNetConstr = NULL,
cores = 4,
verbose = TRUE,
logFile = "log.txt",
seed = NULL,
fileLoadAssoPerm = NULL,
fileLoadCountsPerm = NULL,
storeAssoPerm = FALSE,
fileStoreAssoPerm = "assoPerm",
storeCountsPerm = FALSE,
fileStoreCountsPerm = c("countsPerm1",
"countsPerm2"),
assoPerm = NULL) {
measure <- paramsNetConstruct$measure
measurePar <- paramsNetConstruct$measurePar
zeroMethod <- paramsNetConstruct$zeroMethod
zeroPar <- paramsNetConstruct$zeroPar
needfrac <- paramsNetConstruct$needfrac
needint <- paramsNetConstruct$needint
normMethod <- paramsNetConstruct$normMethod
normPar <- paramsNetConstruct$normPar
jointPrepro <- paramsNetConstruct$jointPrepro
if (!is.null(seed)) {
set.seed(seed)
}
if (jointPrepro) {
n1 <- nrow(normCounts1)
n2 <- nrow(normCounts2)
xbind <- rbind(normCounts1, normCounts2)
} else {
n1 <- nrow(countMat1)
n2 <- nrow(countMat2)
xbind <- rbind(countMat1, countMat2)
}
n <- n1 + n2
nVar <- ncol(assoMat1)
#_____________________________________________________
if (!is.null(fileLoadAssoPerm)) {
stopifnot(is.character(fileLoadAssoPerm))
stopifnot(length(fileLoadAssoPerm) == 1)
fmat <- fm.open(filenamebase = fileLoadAssoPerm,
readonly = TRUE)
if (!(dim(fmat)[1] == nVar * nPerm && dim(fmat)[2] == nVar * 2)) {
stop("fileLoadAssoPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadAssoPerm').")
}
close(fmat)
} else if (!is.null(fileLoadCountsPerm)) {
stopifnot(is.character(fileLoadCountsPerm))
stopifnot(length(fileLoadCountsPerm) == 2)
fmat_counts1 <- fm.open(filenamebase = fileLoadCountsPerm[1],
readonly = TRUE)
fmat_counts2 <- fm.open(filenamebase = fileLoadCountsPerm[2],
readonly = TRUE)
if (!(nrow(fmat_counts1) == nPerm * n1 &&
ncol(fmat_counts1) == nVar)) {
stop("fileLoadCountsPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadCountsPerm').")
}
if (!(nrow(fmat_counts2) == nPerm * n2 &&
ncol(fmat_counts2) == nVar)) {
stop("fileLoadCountsPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadCountsPerm').")
}
close(fmat_counts1)
close(fmat_counts2)
} else {
perm_group_mat <- .getPermGroupMat(n1 = n1, n2 = n2, n = n, nPerm = nPerm,
matchDesign = matchDesign)
}
#-----------------
if (storeCountsPerm) {
stopifnot(is.character(fileStoreCountsPerm))
stopifnot(length(fileStoreCountsPerm) == 2)
fmat_counts1 <- fm.create(filenamebase = fileStoreCountsPerm[1],
nrow = (n1 * nPerm), ncol = nVar)
fmat_counts2 <- fm.create(filenamebase = fileStoreCountsPerm[2],
nrow = (n2 * nPerm), ncol = nVar)
if (verbose) {
message("Files '",
paste0(fileStoreCountsPerm[1], ".bmat, "),
paste0(fileStoreCountsPerm[1], ".desc.txt, "),
paste0(fileStoreCountsPerm[2], ".bmat, and "),
paste0(fileStoreCountsPerm[2], ".desc.txt created. "))
}
close(fmat_counts1)
close(fmat_counts2)
}
#-----------------
if (storeAssoPerm) {
stopifnot(is.character(fileStoreAssoPerm))
stopifnot(length(fileStoreAssoPerm) == 1)
fmat = fm.create(filenamebase = fileStoreAssoPerm,
nrow = (nVar * nPerm), ncol = (2 * nVar))
if (verbose) {
message("Files '",
paste0(fileStoreAssoPerm, ".bmat and "),
paste0(fileStoreAssoPerm, ".desc.txt created. "))
}
close(fmat)
}
#_____________________________________________________
# generate teststatistics for permutated data
if (!is.null(seed)) {
seeds <- sample.int(1e8, size = nPerm)
}
if (verbose) message("Execute permutation tests ... ")
if (cores > 1) {
cl <- parallel::makeCluster(cores, outfile = "")
doSNOW::registerDoSNOW(cl)
if (!is.null(logFile)) cat("", file=logFile, append=FALSE)
'%do_or_dopar%' <- get('%dopar%')
} else {
'%do_or_dopar%' <- get('%do%')
}
if (verbose) {
pb <- utils::txtProgressBar(0, nPerm, style=3)
progress <- function(n) {
utils::setTxtProgressBar(pb,n)
}
opts <- list(progress=progress)
} else {
opts <- list()
}
p <- NULL
#-----------------------------------------------------------------------------
# Run foreach
result <- foreach(
p = 1:nPerm,
.packages = c("filematrix"),
.export = c(
".calcAssociation",
"cclasso",
"gcoda",
".diffConnectPairs",
".diffConnectVariables",
".diffConnectNetwork",
".getVecNames"
),
.options.snow = opts
) %do_or_dopar% {
if (!is.null(logFile)) {
cat(paste("Iteration", p,"\n"),
file=logFile, append=TRUE)
}
if (verbose) progress(p)
if (!is.null(seed)) set.seed(seeds[p])
if (!is.null(assoPerm)) {
# load permutation association matrices (old version)
assoMat1.tmp <- assoPerm[[1]][[p]]
assoMat2.tmp <- assoPerm[[2]][[p]]
count1.tmp <- count2.tmp <- NULL
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
} else if (!is.null(fileLoadAssoPerm )) {
fmat <- fm.open(filenamebase = fileLoadAssoPerm,
readonly = TRUE)
# load permutation asso/diss matrices
assoMat1.tmp <- fmat[(p-1) * nVar + (1:nVar),
1:nVar]
assoMat2.tmp <- fmat[(p-1) * nVar + (1:nVar),
nVar + (1:nVar)]
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
count1.tmp <- count2.tmp <- NULL
close(fmat)
} else {
if (!is.null(fileLoadCountsPerm)) {
fmat_counts1 <- fm.open(filenamebase =
fileLoadCountsPerm[1],
readonly = TRUE)
fmat_counts2 <- fm.open(filenamebase =
fileLoadCountsPerm[2],
readonly = TRUE)
# load permutation count matrices
count1.tmp <-
fmat_counts1[(p-1) * n1 + (1:n1), 1:nVar]
count2.tmp <-
fmat_counts2[(p-1) * n2 + (1:n2), 1:nVar]
close(fmat_counts1)
close(fmat_counts2)
} else {
# generate permutation count matrices
count1.tmp <-
xbind[which(perm_group_mat[p, ] == 1), ]
count2.tmp <-
xbind[which(perm_group_mat[p, ] == 2), ]
if (!jointPrepro) {
# zero treatment and normalization necessary if
# in network construction two count matrices
# were given or dissimilarity network is created
suppressMessages(
count1.tmp <- .zeroTreat(countMat = count1.tmp,
zeroMethod = zeroMethod,
zeroParam = zeroPar,
needfrac = needfrac,
needint = needint,
verbose = FALSE))
suppressMessages(
count2.tmp <- .zeroTreat(countMat = count2.tmp,
zeroMethod = zeroMethod,
zeroParam = zeroPar,
needfrac = needfrac,
needint = needint,
verbose = FALSE))
suppressMessages(
count1.tmp <- .normCounts(countMat = count1.tmp,
normMethod = normMethod,
normParam = normPar,
zeroMethod = zeroMethod,
needfrac = needfrac,
verbose = FALSE))
suppressMessages(
count2.tmp <- .normCounts(countMat = count2.tmp,
normMethod = normMethod,
normParam = normPar,
zeroMethod = zeroMethod,
needfrac = needfrac,
verbose = FALSE))
}
if (storeCountsPerm) {
fmat_counts1 <- fm.open(filenamebase =
fileStoreCountsPerm[1])
fmat_counts2 <- fm.open(filenamebase =
fileStoreCountsPerm[2])
fmat_counts1[(p-1) * n1 + (1:n1),
1:nVar] <- count1.tmp
fmat_counts2[(p-1) * n2 + (1:n2),
1:nVar] <- count2.tmp
close(fmat_counts1)
close(fmat_counts2)
}
}
assoMat1.tmp <- .calcAssociation(count1.tmp,
measure = measure,
measurePar = measurePar,
verbose = FALSE)
assoMat2.tmp <- .calcAssociation(count2.tmp,
measure = measure,
measurePar = measurePar,
verbose = FALSE)
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
if (storeAssoPerm) {
fmat <- fm.open(filenamebase = fileStoreAssoPerm)
fmat[(p-1) * nVar + (1:nVar),
1:nVar] <- assoMat1.tmp
fmat[(p-1) * nVar + (1:nVar),
nVar + (1:nVar)] <- assoMat2.tmp
close(fmat)
}
}
returnlist <- list()
# teststatistics for simulated data
if ("connect.pairs" %in% method) {
connectPairs <- .diffConnectPairs(assoMat1.tmp,
assoMat2.tmp,
fisherTrans)
returnlist[["connectPairs"]] <- connectPairs
}
if ("connect.variables" %in% method) {
connectVariables <- .diffConnectVariables(assoMat1.tmp,
assoMat2.tmp,
nVar,
fisherTrans)
returnlist[["connectVariables"]] <- connectVariables
}
if ("connect.network" %in% method) {
connectNetwork <- .diffConnectNetwork(assoMat1.tmp,
assoMat2.tmp,
nVar,
fisherTrans)
returnlist[["connectNetwork"]] <- connectNetwork
}
returnlist
}
#-----------------------------------------------------------------------------
if (cores > 1) parallel::stopCluster(cl)
#____________________________________________________
output <- list()
if (verbose) close(pb)
# Define whether permutation test is exact or approximative
if (is.null(matchDesign)) {
if (nPerm == choose(n, n1)) {
exact <- TRUE
} else {
exact <- FALSE
}
} else {
if (nPerm == .getMaxCombMatch(matchDesign = matchDesign, n = n)) {
exact <- TRUE
} else {
exact <- FALSE
}
}
if ("connect.pairs" %in% method) {
# test statistic for original data
connectPairsOrig <- .diffConnectPairs(assoMat1, assoMat2, fisherTrans)
# test statistics for simulated data
connectPairs <- matrix(NA, nPerm, length(connectPairsOrig),
dimnames = list(1:nPerm, names(connectPairsOrig)))
for (i in 1:nPerm) {
connectPairs[i, ] <- result[[i]]$connectPairs
}
if (pvalsMethod == "pseudo") {
if (exact) {
# Exact permutation test
pvalsVec <- sapply(1:ncol(connectPairs), function(i) {
sum(connectPairs[, i] >= connectPairsOrig[i]) / nPerm
})
} else {
# Approximated p-values
pvalsVec <- sapply(1:ncol(connectPairs), function(i) {
(sum(connectPairs[, i] >= connectPairsOrig[i]) + 1) / (nPerm + 1)
})
}
} #else {
# pvalsVec <- calc_perm_pvals(tstatPerm = connectPairs,
# tstatObs = connectPairsOrig,
# nExceed = nExceed, ADalpha = 0.05)
# }
names(pvalsVec) <- names(connectPairsOrig)
output[["pvalsVec"]] <- pvalsVec
# adjust for multiple testing
if (verbose & adjust != "none") {
message("Adjust for multiple testing using '", adjust, "' ... ",
appendLF = FALSE)
}
pAdjust <- multAdjust(pvals = pvalsVec, adjust = adjust,
trueNullMethod = trueNullMethod, verbose = verbose)
if (verbose & adjust != "none") message("Done.")
output[["pAdjustVec"]] <- pAdjust
mat.tmp <- assoMat1
mat.tmp[lower.tri(mat.tmp)] <- pvalsVec
mat.tmp[upper.tri(mat.tmp)] <- t(mat.tmp)[upper.tri(t(mat.tmp))]
pvalsMat <- mat.tmp
output[["pvalsMat"]] <- pvalsMat
if (adjust == "none") {
output[["pAdjustMat"]] <- NULL
} else {
mat.tmp[lower.tri(mat.tmp)] <- output[["pAdjustVec"]]
mat.tmp[upper.tri(mat.tmp)] <- t(mat.tmp)[upper.tri(t(mat.tmp))]
output[["pAdjustMat"]] <- mat.tmp
}
output[["testStatData"]] <- connectPairsOrig
output[["testStatPerm"]] <- connectPairs
}
if ("connect.variables" %in% method) {
connectVariablesOrig <- .diffConnectVariables(assoMat1, assoMat2, nVar,
fisherTrans)
connectVariables <- matrix(NA, nPerm, length(connectVariablesOrig),
dimnames = list(1:nPerm,
names(connectVariablesOrig)))
for (i in 1:nPerm) {
connectVariables[i, ] <- result[[i]]$connectVariables
}
if (exact) {
# Exact permutation test
pvalsConnectVariables <-
sapply(1:ncol(connectVariables), function(i) {
sum(connectVariables[, i] >= connectVariablesOrig[i]) / nPerm
})
} else {
# Approximated p-values
pvalsConnectVariables <-
sapply(1:ncol(connectVariables), function(i) {
(sum(connectVariables[, i] >= connectVariablesOrig[i]) + 1) /
(nPerm + 1)
})
}
names(pvalsConnectVariables) <- names(connectVariablesOrig)
output[["pvalsConnectVariables"]] <- pvalsConnectVariables
# adjust for multiple testing
if (adjust2 == "none") {
output[["pAdjustConnectVariables"]] <- NULL
} else if (adjust == "lfdr") {
lfdr <- fdrtool::fdrtool(pvalsConnectVariables, statistic = "pvalue",
plot = FALSE)$lfdr
output[["pAdjustConnectVariables"]] <- lfdr
} else {
p.adj <- p.adjust(pvalsConnectVariables, adjust2)
output[["pAdjustConnectVariables"]] <- p.adj
}
}
if ("connect.network" %in% method) {
connectNetworkOrig <- .diffConnectNetwork(assoMat1, assoMat2, nVar,
fisherTrans)
connectNetwork <- numeric(nPerm)
for (i in 1:nPerm) {
connectNetwork[i] <- result[[i]]$connectNetwork
}
if (exact) {
# Exact permutation test
pvalConnectNetwork <- sum(connectNetwork >= connectNetworkOrig)/ nPerm
} else {
# Approximated p-values
pvalConnectNetwork <- (sum(connectNetwork >= connectNetworkOrig) + 1) /
(nPerm + 1)
}
output[["pvalConnectNetwork"]] <- pvalConnectNetwork
}
return(output)
}
# calculate test statistic for differential connectivity for all variable pairs
.diffConnectPairs <- function(assoMat1, assoMat2, fisherTrans = TRUE) {
# transform distance matrix to vector
diag <- lower.tri(assoMat1, diag = FALSE)
distvec1 <- assoMat1[diag]
distvec2 <- assoMat2[diag]
names(distvec1) <- names(distvec2) <- .getVecNames(assoMat1)
if (fisherTrans) {
# Fisher transformation of correlation coefficients
z1 <- atanh(distvec1)
z2 <- atanh(distvec2)
diff <- abs(z1 - z2)
} else {
diff <- abs(distvec1 - distvec2)
}
return(diff)
}
# calculate test statistic for difference in connectivity between a single
# variable and all other variables
.diffConnectVariables <- function(assoMat1, assoMat2, nVar, fisherTrans = TRUE) {
if (fisherTrans) {
# Fisher transformation of correlation coefficients
Z1 <- atanh(assoMat1)
Z2 <- atanh(assoMat2)
# build matrix with absolute differences of z-values
diff <- abs(Z1 - Z2)
diag(diff) <- 0
} else {
diff <- abs(assoMat1 - assoMat2)
}
### test statistic
d_vec <- numeric(nVar)
# calculate test statistics for all variables
for (i in 1:nVar) {
d_vec[i] <- sum(diff[i, -i]) / (nVar - 1)
}
names(d_vec) <- colnames(assoMat1)
return(d_vec)
}
# calculate test statistic for difference in connectivity for the whole network
.diffConnectNetwork <- function(assoMat1, assoMat2, nVar, fisherTrans = TRUE) {
if (fisherTrans) {
# Fisher transformation of correlation coefficients
Z1 <- atanh(assoMat1)
Z2 <- atanh(assoMat2)
# build matrix with absolute differences of z-values
diff <- abs(Z1 - Z2)
diag(diff) <- 0
} else {
diff <- abs(assoMat1 - assoMat2)
}
# test statistic
d <- (sum(diff) - sum(diag(diff))) / (nVar * (nVar-1))
return(d)
}
stefpeschel/NetCoMi documentation built on Feb. 4, 2024, 8:20 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions .diffConnectNetwork .diffConnectVariables .diffConnectPairs .permTestDiffAsso
Documented in .permTestDiffAsso
#' @title Permutation Tests for Determining Differential Associations
#'
#' @description The function implements procedures to test whether pairs of taxa
#' are differentially associated, whether a taxon is differentially associated
#' to all other taxa, or whether two networks are differentially associated
#' between two groups as proposed by \cite{Gill et al.(2010)}.
#'
#' @param countMat1,countMat2 matrices containing microbiome data (read counts)
#' of group 1 and group 2 (rows represent samples and columns taxonomic units,
#' respectively).
#' @param countsJoint joint count matrices before preprocessing
#' @param normCounts1,normCounts2 normalized count matrices.
#' @param assoMat1,assoMat2 association matrices corresponding to the two count
#' matrices. The associations must have been estimated from the count matrices
#' \code{countMat1} and \code{countMat2}.
#' @param paramsNetConstruct parameters used for network construction.
#' @param method character vector indicating the tests to be performed. Possible
#' values are \code{"connect.pairs"} (differentially correlated taxa pairs),
#' \code{"connect.variables"} (one taxon to all other) and
#' \code{"connect.network"} (differentially connected networks). By default,
#' all three tests are conducted.
#' @param fisherTrans logical indicating whether the correlation values should
#' be Fisher-transformed.
#' @param pvalsMethod currently only \code{"pseudo"} is available, where 1 is
#' added to the number of permutations and the permutation test statistics
#' being more extreme than the observed one in order to avoid zero p-values.
#' @param adjust multiple testing adjustment for the tests for differentially
#' correlated pairs of taxa; possible values are "lfdr" (default) for local
#' false discovery rate correction (via \code{\link[fdrtool]{fdrtool}}) or one
#' of the methods provided by \code{\link[stats]{p.adjust}}
#' @param adjust2 multiple testing adjustment for the tests if a taxa pair is
#' differentially correlated to all other taxa; possible methods are those
#' provided by \code{\link[stats]{p.adjust}} (a few hundred tests are
#' necessary for the local fdr correction)
#' @param trueNullMethod character indicating the method used for estimating the
#' proportion of true null hypotheses from a vector of p-values. Used for the
#' adaptive Benjamini-Hochberg method for multiple testing adjustment (chosen
#' by \code{adjust = "adaptBH"}).
#' @param alpha significance level
#' @param lfdrThresh defines a threshold for the local fdr if "lfdr" is chosen
#' as method for multiple testing correction; defaults to 0.2, which means
#' that correlations with a corresponding local fdr less than or equal to 0.2
#' are identified as significant
#' @param nPerm number of permutations
#' @param matchDesign Numeric vector with two elements specifying an optional
#' matched-group (i.e. matched-pair) design, which is used for the permutation
#' tests in \code{\link{netCompare}} and \code{\link{diffnet}}. \code{c(1,1)}
#' corresponds to a matched-pair design. A 1:2 matching, for instance, is
#' defined by \code{c(1,2)}, which means that the first sample of group 1 is
#' matched to the first two samples of group 2 and so on.
#' The appropriate order of samples must be ensured. If
#' \code{NULL}, the group memberships are shuffled randomly while group sizes
#' identical to the original data set are ensured.
#' @param callNetConstr call inherited from \code{netConstruct()}.
#' @param cores number of CPU cores (permutation tests are executed parallel)
#' @param verbose if \code{TRUE}, status messages and numbers of SparCC
#' iterations are printed
#' @param logFile character string naming the log file within which the current
#' iteration number is stored
#' @param seed an optional seed for reproducibility of the results
#' @param fileLoadAssoPerm character giving the name (without extenstion)
#' or path of the file storing the "permuted" association/dissimilarity
#' matrices that have been exported by setting \code{storeAssoPerm} to
#' \code{TRUE}. Only used for permutation tests. Set to \code{NULL} if no
#' existing associations should be used.
#' @param fileLoadCountsPerm character giving the name (without extenstion)
#' or path of the file storing the "permuted" count matrices that have been
#' exported by setting \code{storeCountsPerm} to \code{TRUE}.
#' Only used for permutation tests, and if \code{fileLoadAssoPerm = NULL}.
#' Set to \code{NULL} if no existing count matrices should be used.
#' @param storeAssoPerm logical indicating whether the association (or
#' dissimilarity) matrices for the permuted data should be stored in a file.
#' The filename is given via \code{fileStoreAssoPerm}. If \code{TRUE},
#' the computed "permutation" association/dissimilarity matrices can be reused
#' via \code{fileLoadAssoPerm} to save runtime. Defaults to \code{FALSE}.
#' @param fileStoreAssoPerm character giving the file name to store a matrix
#' containing a matrix with associations/dissimilarities for the permuted
#' data. Can also be a path.
#' @param storeCountsPerm logical indicating whether the permuted count matrices
#' should be stored in an external file. Defaults to \code{FALSE}.
#' @param fileStoreCountsPerm character vector with two elements giving the
#' names of two files storing the permuted count matrices belonging to the
#' two groups.
#' @param assoPerm not used anymore.
#' @references
#' \insertRef{gill2010statistical}{NetCoMi}\cr\cr
#' \insertRef{knijnenburg2009fewer}{NetCoMi}\cr\cr
#'
#' @seealso \code{\link{diffnet}}
.permTestDiffAsso <- function(countMat1, countMat2, countsJoint,
normCounts1, normCounts2,
assoMat1, assoMat2,
paramsNetConstruct,
method = c("connect.pairs",
"connect.variables",
"connect.network"),
fisherTrans = TRUE,
pvalsMethod = "pseudo",
adjust = "lfdr",
adjust2 = "holm",
trueNullMethod = "convest",
alpha = 0.05,
lfdrThresh = 0.2,
nPerm = 1000,
matchDesign = NULL,
callNetConstr = NULL,
cores = 4,
verbose = TRUE,
logFile = "log.txt",
seed = NULL,
fileLoadAssoPerm = NULL,
fileLoadCountsPerm = NULL,
storeAssoPerm = FALSE,
fileStoreAssoPerm = "assoPerm",
storeCountsPerm = FALSE,
fileStoreCountsPerm = c("countsPerm1",
"countsPerm2"),
assoPerm = NULL) {
measure <- paramsNetConstruct$measure
measurePar <- paramsNetConstruct$measurePar
zeroMethod <- paramsNetConstruct$zeroMethod
zeroPar <- paramsNetConstruct$zeroPar
needfrac <- paramsNetConstruct$needfrac
needint <- paramsNetConstruct$needint
normMethod <- paramsNetConstruct$normMethod
normPar <- paramsNetConstruct$normPar
jointPrepro <- paramsNetConstruct$jointPrepro
if (!is.null(seed)) {
set.seed(seed)
}
if (jointPrepro) {
n1 <- nrow(normCounts1)
n2 <- nrow(normCounts2)
xbind <- rbind(normCounts1, normCounts2)
} else {
n1 <- nrow(countMat1)
n2 <- nrow(countMat2)
xbind <- rbind(countMat1, countMat2)
}
n <- n1 + n2
nVar <- ncol(assoMat1)
#_____________________________________________________
if (!is.null(fileLoadAssoPerm)) {
stopifnot(is.character(fileLoadAssoPerm))
stopifnot(length(fileLoadAssoPerm) == 1)
fmat <- fm.open(filenamebase = fileLoadAssoPerm,
readonly = TRUE)
if (!(dim(fmat)[1] == nVar * nPerm && dim(fmat)[2] == nVar * 2)) {
stop("fileLoadAssoPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadAssoPerm').")
}
close(fmat)
} else if (!is.null(fileLoadCountsPerm)) {
stopifnot(is.character(fileLoadCountsPerm))
stopifnot(length(fileLoadCountsPerm) == 2)
fmat_counts1 <- fm.open(filenamebase = fileLoadCountsPerm[1],
readonly = TRUE)
fmat_counts2 <- fm.open(filenamebase = fileLoadCountsPerm[2],
readonly = TRUE)
if (!(nrow(fmat_counts1) == nPerm * n1 &&
ncol(fmat_counts1) == nVar)) {
stop("fileLoadCountsPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadCountsPerm').")
}
if (!(nrow(fmat_counts2) == nPerm * n2 &&
ncol(fmat_counts2) == nVar)) {
stop("fileLoadCountsPerm has wrong dimensions. 'nPerm' might be set ",
"incorrectly (must equal the number of permutation matrices in ",
"'fileLoadCountsPerm').")
}
close(fmat_counts1)
close(fmat_counts2)
} else {
perm_group_mat <- .getPermGroupMat(n1 = n1, n2 = n2, n = n, nPerm = nPerm,
matchDesign = matchDesign)
}
#-----------------
if (storeCountsPerm) {
stopifnot(is.character(fileStoreCountsPerm))
stopifnot(length(fileStoreCountsPerm) == 2)
fmat_counts1 <- fm.create(filenamebase = fileStoreCountsPerm[1],
nrow = (n1 * nPerm), ncol = nVar)
fmat_counts2 <- fm.create(filenamebase = fileStoreCountsPerm[2],
nrow = (n2 * nPerm), ncol = nVar)
if (verbose) {
message("Files '",
paste0(fileStoreCountsPerm[1], ".bmat, "),
paste0(fileStoreCountsPerm[1], ".desc.txt, "),
paste0(fileStoreCountsPerm[2], ".bmat, and "),
paste0(fileStoreCountsPerm[2], ".desc.txt created. "))
}
close(fmat_counts1)
close(fmat_counts2)
}
#-----------------
if (storeAssoPerm) {
stopifnot(is.character(fileStoreAssoPerm))
stopifnot(length(fileStoreAssoPerm) == 1)
fmat = fm.create(filenamebase = fileStoreAssoPerm,
nrow = (nVar * nPerm), ncol = (2 * nVar))
if (verbose) {
message("Files '",
paste0(fileStoreAssoPerm, ".bmat and "),
paste0(fileStoreAssoPerm, ".desc.txt created. "))
}
close(fmat)
}
#_____________________________________________________
# generate teststatistics for permutated data
if (!is.null(seed)) {
seeds <- sample.int(1e8, size = nPerm)
}
if (verbose) message("Execute permutation tests ... ")
if (cores > 1) {
cl <- parallel::makeCluster(cores, outfile = "")
doSNOW::registerDoSNOW(cl)
if (!is.null(logFile)) cat("", file=logFile, append=FALSE)
'%do_or_dopar%' <- get('%dopar%')
} else {
'%do_or_dopar%' <- get('%do%')
}
if (verbose) {
pb <- utils::txtProgressBar(0, nPerm, style=3)
progress <- function(n) {
utils::setTxtProgressBar(pb,n)
}
opts <- list(progress=progress)
} else {
opts <- list()
}
p <- NULL
#-----------------------------------------------------------------------------
# Run foreach
result <- foreach(
p = 1:nPerm,
.packages = c("filematrix"),
.export = c(
".calcAssociation",
"cclasso",
"gcoda",
".diffConnectPairs",
".diffConnectVariables",
".diffConnectNetwork",
".getVecNames"
),
.options.snow = opts
) %do_or_dopar% {
if (!is.null(logFile)) {
cat(paste("Iteration", p,"\n"),
file=logFile, append=TRUE)
}
if (verbose) progress(p)
if (!is.null(seed)) set.seed(seeds[p])
if (!is.null(assoPerm)) {
# load permutation association matrices (old version)
assoMat1.tmp <- assoPerm[[1]][[p]]
assoMat2.tmp <- assoPerm[[2]][[p]]
count1.tmp <- count2.tmp <- NULL
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
} else if (!is.null(fileLoadAssoPerm )) {
fmat <- fm.open(filenamebase = fileLoadAssoPerm,
readonly = TRUE)
# load permutation asso/diss matrices
assoMat1.tmp <- fmat[(p-1) * nVar + (1:nVar),
1:nVar]
assoMat2.tmp <- fmat[(p-1) * nVar + (1:nVar),
nVar + (1:nVar)]
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
count1.tmp <- count2.tmp <- NULL
close(fmat)
} else {
if (!is.null(fileLoadCountsPerm)) {
fmat_counts1 <- fm.open(filenamebase =
fileLoadCountsPerm[1],
readonly = TRUE)
fmat_counts2 <- fm.open(filenamebase =
fileLoadCountsPerm[2],
readonly = TRUE)
# load permutation count matrices
count1.tmp <-
fmat_counts1[(p-1) * n1 + (1:n1), 1:nVar]
count2.tmp <-
fmat_counts2[(p-1) * n2 + (1:n2), 1:nVar]
close(fmat_counts1)
close(fmat_counts2)
} else {
# generate permutation count matrices
count1.tmp <-
xbind[which(perm_group_mat[p, ] == 1), ]
count2.tmp <-
xbind[which(perm_group_mat[p, ] == 2), ]
if (!jointPrepro) {
# zero treatment and normalization necessary if
# in network construction two count matrices
# were given or dissimilarity network is created
suppressMessages(
count1.tmp <- .zeroTreat(countMat = count1.tmp,
zeroMethod = zeroMethod,
zeroParam = zeroPar,
needfrac = needfrac,
needint = needint,
verbose = FALSE))
suppressMessages(
count2.tmp <- .zeroTreat(countMat = count2.tmp,
zeroMethod = zeroMethod,
zeroParam = zeroPar,
needfrac = needfrac,
needint = needint,
verbose = FALSE))
suppressMessages(
count1.tmp <- .normCounts(countMat = count1.tmp,
normMethod = normMethod,
normParam = normPar,
zeroMethod = zeroMethod,
needfrac = needfrac,
verbose = FALSE))
suppressMessages(
count2.tmp <- .normCounts(countMat = count2.tmp,
normMethod = normMethod,
normParam = normPar,
zeroMethod = zeroMethod,
needfrac = needfrac,
verbose = FALSE))
}
if (storeCountsPerm) {
fmat_counts1 <- fm.open(filenamebase =
fileStoreCountsPerm[1])
fmat_counts2 <- fm.open(filenamebase =
fileStoreCountsPerm[2])
fmat_counts1[(p-1) * n1 + (1:n1),
1:nVar] <- count1.tmp
fmat_counts2[(p-1) * n2 + (1:n2),
1:nVar] <- count2.tmp
close(fmat_counts1)
close(fmat_counts2)
}
}
assoMat1.tmp <- .calcAssociation(count1.tmp,
measure = measure,
measurePar = measurePar,
verbose = FALSE)
assoMat2.tmp <- .calcAssociation(count2.tmp,
measure = measure,
measurePar = measurePar,
verbose = FALSE)
dimnames(assoMat1.tmp) <- dimnames(assoMat1)
dimnames(assoMat2.tmp) <- dimnames(assoMat2)
if (storeAssoPerm) {
fmat <- fm.open(filenamebase = fileStoreAssoPerm)
fmat[(p-1) * nVar + (1:nVar),
1:nVar] <- assoMat1.tmp
fmat[(p-1) * nVar + (1:nVar),
nVar + (1:nVar)] <- assoMat2.tmp
close(fmat)
}
}
returnlist <- list()
# teststatistics for simulated data
if ("connect.pairs" %in% method) {
connectPairs <- .diffConnectPairs(assoMat1.tmp,
assoMat2.tmp,
fisherTrans)
returnlist[["connectPairs"]] <- connectPairs
}
if ("connect.variables" %in% method) {
connectVariables <- .diffConnectVariables(assoMat1.tmp,
assoMat2.tmp,
nVar,
fisherTrans)
returnlist[["connectVariables"]] <- connectVariables
}
if ("connect.network" %in% method) {
connectNetwork <- .diffConnectNetwork(assoMat1.tmp,
assoMat2.tmp,
nVar,
fisherTrans)
returnlist[["connectNetwork"]] <- connectNetwork
}
returnlist
}
#-----------------------------------------------------------------------------
if (cores > 1) parallel::stopCluster(cl)
#____________________________________________________
output <- list()
if (verbose) close(pb)
# Define whether permutation test is exact or approximative
if (is.null(matchDesign)) {
if (nPerm == choose(n, n1)) {
exact <- TRUE
} else {
exact <- FALSE
}
} else {
if (nPerm == .getMaxCombMatch(matchDesign = matchDesign, n = n)) {
exact <- TRUE
} else {
exact <- FALSE
}
}
if ("connect.pairs" %in% method) {
# test statistic for original data
connectPairsOrig <- .diffConnectPairs(assoMat1, assoMat2, fisherTrans)
# test statistics for simulated data
connectPairs <- matrix(NA, nPerm, length(connectPairsOrig),
dimnames = list(1:nPerm, names(connectPairsOrig)))
for (i in 1:nPerm) {
connectPairs[i, ] <- result[[i]]$connectPairs
}
if (pvalsMethod == "pseudo") {
if (exact) {
# Exact permutation test
pvalsVec <- sapply(1:ncol(connectPairs), function(i) {
sum(connectPairs[, i] >= connectPairsOrig[i]) / nPerm
})
} else {
# Approximated p-values
pvalsVec <- sapply(1:ncol(connectPairs), function(i) {
(sum(connectPairs[, i] >= connectPairsOrig[i]) + 1) / (nPerm + 1)
})
}
} #else {
# pvalsVec <- calc_perm_pvals(tstatPerm = connectPairs,
# tstatObs = connectPairsOrig,
# nExceed = nExceed, ADalpha = 0.05)
# }
names(pvalsVec) <- names(connectPairsOrig)
output[["pvalsVec"]] <- pvalsVec
# adjust for multiple testing
if (verbose & adjust != "none") {
message("Adjust for multiple testing using '", adjust, "' ... ",
appendLF = FALSE)
}
pAdjust <- multAdjust(pvals = pvalsVec, adjust = adjust,
trueNullMethod = trueNullMethod, verbose = verbose)
if (verbose & adjust != "none") message("Done.")
output[["pAdjustVec"]] <- pAdjust
mat.tmp <- assoMat1
mat.tmp[lower.tri(mat.tmp)] <- pvalsVec
mat.tmp[upper.tri(mat.tmp)] <- t(mat.tmp)[upper.tri(t(mat.tmp))]
pvalsMat <- mat.tmp
output[["pvalsMat"]] <- pvalsMat
if (adjust == "none") {
output[["pAdjustMat"]] <- NULL
} else {
mat.tmp[lower.tri(mat.tmp)] <- output[["pAdjustVec"]]
mat.tmp[upper.tri(mat.tmp)] <- t(mat.tmp)[upper.tri(t(mat.tmp))]
output[["pAdjustMat"]] <- mat.tmp
}
output[["testStatData"]] <- connectPairsOrig
output[["testStatPerm"]] <- connectPairs
}
if ("connect.variables" %in% method) {
connectVariablesOrig <- .diffConnectVariables(assoMat1, assoMat2, nVar,
fisherTrans)
connectVariables <- matrix(NA, nPerm, length(connectVariablesOrig),
dimnames = list(1:nPerm,
names(connectVariablesOrig)))
for (i in 1:nPerm) {
connectVariables[i, ] <- result[[i]]$connectVariables
}
if (exact) {
# Exact permutation test
pvalsConnectVariables <-
sapply(1:ncol(connectVariables), function(i) {
sum(connectVariables[, i] >= connectVariablesOrig[i]) / nPerm
})
} else {
# Approximated p-values
pvalsConnectVariables <-
sapply(1:ncol(connectVariables), function(i) {
(sum(connectVariables[, i] >= connectVariablesOrig[i]) + 1) /
(nPerm + 1)
})
}
names(pvalsConnectVariables) <- names(connectVariablesOrig)
output[["pvalsConnectVariables"]] <- pvalsConnectVariables
# adjust for multiple testing
if (adjust2 == "none") {
output[["pAdjustConnectVariables"]] <- NULL
} else if (adjust == "lfdr") {
lfdr <- fdrtool::fdrtool(pvalsConnectVariables, statistic = "pvalue",
plot = FALSE)$lfdr
output[["pAdjustConnectVariables"]] <- lfdr
} else {
p.adj <- p.adjust(pvalsConnectVariables, adjust2)
output[["pAdjustConnectVariables"]] <- p.adj
}
}
if ("connect.network" %in% method) {
connectNetworkOrig <- .diffConnectNetwork(assoMat1, assoMat2, nVar,
fisherTrans)
connectNetwork <- numeric(nPerm)
for (i in 1:nPerm) {
connectNetwork[i] <- result[[i]]$connectNetwork
}
if (exact) {
# Exact permutation test
pvalConnectNetwork <- sum(connectNetwork >= connectNetworkOrig)/ nPerm
} else {
# Approximated p-values
pvalConnectNetwork <- (sum(connectNetwork >= connectNetworkOrig) + 1) /
(nPerm + 1)
}
output[["pvalConnectNetwork"]] <- pvalConnectNetwork
}
return(output)
}
# calculate test statistic for differential connectivity for all variable pairs
.diffConnectPairs <- function(assoMat1, assoMat2, fisherTrans = TRUE) {
# transform distance matrix to vector
diag <- lower.tri(assoMat1, diag = FALSE)
distvec1 <- assoMat1[diag]
distvec2 <- assoMat2[diag]
names(distvec1) <- names(distvec2) <- .getVecNames(assoMat1)
if (fisherTrans) {
# Fisher transformation of correlation coefficients
z1 <- atanh(distvec1)
z2 <- atanh(distvec2)
diff <- abs(z1 - z2)
} else {
diff <- abs(distvec1 - distvec2)
}
return(diff)
}
# calculate test statistic for difference in connectivity between a single
# variable and all other variables
.diffConnectVariables <- function(assoMat1, assoMat2, nVar, fisherTrans = TRUE) {
if (fisherTrans) {
# Fisher transformation of correlation coefficients
Z1 <- atanh(assoMat1)
Z2 <- atanh(assoMat2)
# build matrix with absolute differences of z-values
diff <- abs(Z1 - Z2)
diag(diff) <- 0
} else {
diff <- abs(assoMat1 - assoMat2)
}
### test statistic
d_vec <- numeric(nVar)
# calculate test statistics for all variables
for (i in 1:nVar) {
d_vec[i] <- sum(diff[i, -i]) / (nVar - 1)
}
names(d_vec) <- colnames(assoMat1)
return(d_vec)
}
# calculate test statistic for difference in connectivity for the whole network
.diffConnectNetwork <- function(assoMat1, assoMat2, nVar, fisherTrans = TRUE) {
if (fisherTrans) {
# Fisher transformation of correlation coefficients
Z1 <- atanh(assoMat1)
Z2 <- atanh(assoMat2)
# build matrix with absolute differences of z-values
diff <- abs(Z1 - Z2)
diag(diff) <- 0
} else {
diff <- abs(assoMat1 - assoMat2)
}
# test statistic
d <- (sum(diff) - sum(diag(diff))) / (nVar * (nVar-1))
return(d)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.