R/globalstd.R
In yufree/pmd: Paired Mass Distance Analysis for GC/LC-MS Based Non-Targeted Analysis and Reactomics Analysis
Defines functions
pcasf
getcluster
getcorcluster
globalstd
getstd
getpaired
Documented in
getcluster
getcorcluster
getpaired
getstd
globalstd
pcasf
#' Filter ions/peaks based on retention time hierarchical clustering, paired mass distances(PMD) and PMD frequency analysis.
#' @param list a list with mzrt profile
#' @param rtcutoff cutoff of the distances in retention time hierarchical clustering analysis, default 10
#' @param ng cutoff of global PMD's retention time group numbers, If ng = NULL, 20 percent of RT cluster will be used as ng, default NULL.
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @param mdrange mass defect range to ignore. Default NULL and c(0.25,0.9) to retain the possible reaction related paired mass
#' @return list with tentative isotope, multi-chargers, adducts, and neutral loss peaks' index, retention time clusters.
#' @examples
#' data(spmeinvivo)
#' pmd <- getpaired(spmeinvivo)
#' @seealso \code{\link{getstd}},\code{\link{getsda}},\code{\link{plotpaired}}
#' @export
getpaired <-
function(list,
rtcutoff = 10,
ng = NULL,
digits = 2,
accuracy = 4,
mdrange = NULL) {
# Convert a pmd data.frame to isotope, multi, and diff pmd data.frame list.
rtpmd <- function(bin, i) {
medianrtxi <- stats::median(bin$rt)
if (ncol(bin) > 2) {
if (nrow(bin) > 1) {
# get mz diff
dis <-
stats::dist(bin$mz, method = "manhattan")
# get intensity cor
cor <-
stats::cor(t(bin[, -c(1, 2)]))
df <-
data.frame(
ms1 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = as.numeric(dis),
rt = medianrtxi,
rtg = i,
cor = cor[lower.tri(cor)]
)
# remove isotope
isoindex <-
(round(df$diff, digits) != 0) &
((
df$diff %% 1 < 0.01 &
df$diff >= 1 &
df$diff < 2
) | (
df$diff %% 2 < 0.01 &
df$diff >= 2 &
df$diff < 3
) | (
df$diff %% 1 > 0.99 &
df$diff >= 1 &
df$diff < 2
) | (
df$diff %% 1 > 0.99 &
df$diff >= 0 &
df$diff < 1
)
)
# higher is C13
# massstd <-
# apply(df[isoindex,], 1, function(x)
# min(x[1], x[2]))
massstdmax <-
apply(df[isoindex,], 1, function(x)
max(x[1], x[2]))
isomass <-
unique(massstdmax)
dfiso <- df[isoindex, ]
if (nrow(dfiso) > 0) {
df <-
df[!(df[, 1] %in% isomass) &
!(df[, 2] %in% isomass), ]
}
# remove multi chargers
multiindex <-
(round(df$diff %% 1, 1) == 0.5)
dfmulti <- df[multiindex, ]
mass <-
unique(df[multiindex, 1], df[multiindex, 2])
# From HMDB lowest mass with 0.5 is 394.4539, remove those ions related pmd
multimass <-
mass[round(mass %% 1, 1) == 0.5 & mass<350]
if (nrow(dfmulti) > 0) {
df <-
df[!(df[, 1] %in% multimass) & !(df[, 2] %in% multimass),]
}
dfdiff <- df
return(
list(
dfmulti = dfmulti,
dfiso = dfiso,
dfdiff = dfdiff,
solo = NULL
)
)
} else {
solo <- cbind(bin[, c(1:2)],
rtg = i,
cor = 1)
return(
list(
dfmulti = NULL,
dfiso = NULL,
dfdiff = NULL,
solo = solo
)
)
}
} else{
if (nrow(bin) > 1) {
# get mz diff
dis <-
stats::dist(bin$mz, method = "manhattan")
df <-
data.frame(
ms1 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = as.numeric(dis),
rt = medianrtxi,
rtg = i
)
# higher is C13
# massstd <-
# apply(df[isoindex,], 1, function(x)
# min(x[1], x[2]))
massstdmax <-
apply(df[isoindex,], 1, function(x)
max(x[1], x[2]))
isomass <-
unique(massstdmax)
dfiso <- df[isoindex, ]
if (nrow(dfiso) > 0) {
df <-
df[!(df[, 1] %in% isomass) &
!(df[, 2] %in% isomass), ]
}
# remove multi chargers
multiindex <-
(round(df$diff %% 1, 1) == 0.5)
dfmulti <- df[multiindex, ]
mass <-
unique(df[multiindex, 1], df[multiindex, 2])
# From HMDB lowest mass with 0.5 is 394.4539, remove those ions related pmd
multimass <-
mass[round(mass %% 1, 1) == 0.5 & mass<350]
if (nrow(dfmulti) > 0) {
df <-
df[!(df[, 1] %in% multimass) & !(df[, 2] %in% multimass),]
}
dfdiff <- df
return(
list(
dfmulti = dfmulti,
dfiso = dfiso,
dfdiff = dfdiff,
solo = NULL
)
)
} else {
solo <- cbind(bin, rtg = i)
return(
list(
dfmulti = NULL,
dfiso = NULL,
dfdiff = NULL,
solo = solo
)
)
}
}
}
# paired mass diff analysis
dis <- stats::dist(list$rt, method = "manhattan")
fit <- stats::hclust(dis)
rtcluster <- stats::cutree(fit, h = rtcutoff)
n <- length(unique(rtcluster))
message(paste(n, "retention time cluster found."))
# automate ng selection when ng is NULL
ng <- ifelse(is.null(ng), round(n * 0.2), ng)
if (!is.null(list$data)) {
groups <- cbind.data.frame(mz = list$mz,
rt = list$rt,
list$data)
} else {
groups <- cbind.data.frame(mz = list$mz,
rt = list$rt)
message('No intensity data!')
}
split <- split.data.frame(groups, rtcluster)
rtpmdtemp <-
mapply(rtpmd,
split,
as.numeric(names(split)),
SIMPLIFY = FALSE)
result <- do.call(Map, c(rbind, rtpmdtemp))
# filter the list get the rt cluster
list$rtcluster <- rtcluster
result$dfdiff$diff2 <-
round(result$dfdiff$diff, digits)
result$dfdiff$md <- result$dfdiff$diff%%1
# speed up by mass defect
if(is.null(mdrange)){
list$diff <- result$dfdiff
}else{
list$diff <- result$dfdiff <- result$dfdiff[result$dfdiff$md<mdrange[1]|result$dfdiff$md>mdrange[2],]
}
# speed up by group
pmd <-
as.numeric(names(table(result$dfdiff$diff2)[table(result$dfdiff$diff2) > ng]))
# pmd <- unique(result$dfdiff$diff2)
idx <- NULL
for (i in seq_along(pmd)) {
l <-
length(unique(result$dfdiff[result$dfdiff$diff2 == pmd[i], 'rtg']))
idx <-
c(idx, ifelse(l > ng, TRUE, FALSE))
}
pmd2 <- pmd[idx]
list$paired <-
result$dfdiff[result$dfdiff$diff2 %in% pmd2, ]
if (nrow(result$dfdiff) > 0) {
list$pairedindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(list$paired$ms1, accuracy),
round(list$paired$ms2,
accuracy)
),
c(list$paired$rtg, list$paired$rtg))
}
# get the data index by rt groups with high frequencies
# PMD
list$diffindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfdiff$ms1, accuracy),
round(result$dfdiff$ms2,
accuracy)
),
c(result$dfdiff$rtg, result$dfdiff$rtg))
# get the data index by rt groups with single ions
if (!is.null(result$solo)) {
list$soloindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(round(result$solo$mz, accuracy),
result$solo$rtg)
list$solo <- result$solo
}
# get the data index by rt groups with isotope ions
if (!is.null(result$dfiso)) {
list$isoindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfiso$ms1, accuracy),
round(result$dfiso$ms2, accuracy)
),
c(result$dfiso$rtg, result$dfiso$rtg))
list$iso <- result$dfiso
}
# get the data index by rt groups with multi charger ions
if (!is.null(result$dfmulti)) {
list$multiindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfmulti$ms1, accuracy),
round(result$dfmulti$ms2,
accuracy)
),
c(result$dfmulti$rtg, result$dfmulti$rtg))
list$multi <- result$dfmulti
}
# show message about std mass
message(paste(sum(list$pairedindex), "paired masses found "))
message(
paste(
length(unique(list$paired$diff2)),
"unique within RT clusters high frequency PMD(s) used for further investigation."
)
)
message(paste(c("The unique within RT clusters high frequency PMD(s) is(are) ", unique(list$paired$diff2)), collapse=" "), '.')
message(paste(
sum(list$isoindex),
"isotopologue(s) related paired mass found."
))
message(paste(
sum(list$multiindex),
"multi-charger(s) related paired mass found."
))
# return results
return(list)
}
#' Find the independent ions for each retention time hierarchical clustering based on PMD relationship within each retention time cluster and isotope and return the index of the std data for each retention time cluster.
#' @param list a list from getpaired function
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with std mass index
#' @examples
#' data(spmeinvivo)
#' pmd <- getpaired(spmeinvivo)
#' std <- getstd(pmd)
#' @seealso \code{\link{getpaired}},\code{\link{getsda}},\code{\link{plotstd}}
#' @export
getstd <-
function(list,
corcutoff = NULL,
digits = 2,
accuracy = 4) {
resultstd2A <- resultstd2B1 <- resultstd2B2 <- resultstd2B3 <- NULL
# filter high freq ions and find std mass
resultdiff <- list$paired
resultiso <- list$iso
if (!is.null(corcutoff)) {
resultdiff <- resultdiff[abs(resultdiff$cor) >= corcutoff,]
resultiso <- resultiso[resultiso$cor >= corcutoff,]
}
# filter the mass from mass pairs within retention time
# group group 1: RT groups with solo peak
resultstd1 <- NULL
if (!is.null(list$solo)) {
resultstd1 <- cbind(list$solo$mz, list$solo$rt,
list$solo$rtg)
}
message(paste(
c(
sum(list$soloindex),
"retention group(s) have single peaks.",
list$rtcluster[list$soloindex]
),
collapse = ' '
))
# group 2: RT groups with multiple peaks group 2A: RT
# groups with multiple peaks while no isotope/paired
# relationship
index2A <-
!(
unique(list$rtcluster) %in% unique(resultdiff$rtg) |
unique(list$rtcluster) %in% unique(resultiso$rtg)
) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2A <- unique(list$rtcluster)[index2A]
message(paste(
c(
sum(index2A),
'group(s) with multiple peaks while no isotope/paired relationship',
rtg2A
),
collapse = " "
))
# print(rtg2A)
if (sum(index2A) > 0) {
for (i in seq_along(rtg2A)) {
mass <- list$mz[list$rtcluster == rtg2A[i]]
rt <- list$rt[list$rtcluster == rtg2A[i]]
mass <- max(mass)
suppressWarnings(resultstdtemp <-
c(mass, stats::median(rt),
rtg2A[i]))
suppressWarnings(resultstd2A <-
rbind(resultstd2A,
resultstdtemp))
}
}
# group 2B: RT groups with multiple peaks with
# isotope/paired relationship index2B <-
# (unique(list$rtcluster) %in%
# unique(resultdiff$rtg))|(unique(list$rtcluster) %in%
# unique(resultiso$rtg)) group 2B1: RT groups with
# multiple peaks with isotope without paired
# relationship
index2B1 <-
!(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
unique(list$rtcluster) %in% unique(resultiso$rtg) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B1 <- unique(list$rtcluster)[index2B1]
# print(rtg2B1)
message(paste(
c(
sum(index2B1),
'group(s) with multiple peaks with isotope without paired relationship',
rtg2B1
),
collapse = " "
))
if (sum(index2B1) > 0) {
for (i in seq_along(rtg2B1)) {
# filter the isotope peaks
dfiso <-
resultiso[resultiso$rtg == rtg2B1[i],]
if (nrow(dfiso) > 0) {
massstd <- apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
mass <-
unique(massstd[!(massstd %in% massstdmax)])
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
resultstd2B1 <-
rbind(resultstd2B1, resultstdtemp)
} else{
resultstd2B1 <- NULL
}
}
}
# group 2B2: RT groups with multiple peaks with paired
# relationship without isotope
index2B2 <-
(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
!(unique(list$rtcluster) %in% unique(resultiso$rtg)) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B2 <- unique(list$rtcluster)[index2B2]
message(paste(
c(
sum(index2B2),
'group(s) with paired relationship without isotope',
rtg2B2
),
collapse = ' '
))
# print(rtg2B2)
if (sum(index2B2) > 0) {
for (i in seq_along(rtg2B2)) {
# filter the paired peaks
df <-
resultdiff[resultdiff$rtg == rtg2B2[i],]
if (nrow(df) > 0) {
mass <- apply(df, 1, function(x)
min(x[1],
x[2]))
mass <- unique(mass)
suppressWarnings(
resultstdtemp <-
cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = df$rt,
rtg = df$rtg
)
)
suppressWarnings(
resultstd2B2 <- rbind(resultstd2B2,
resultstdtemp)
)
}
}
}
# group 2B3: RT groups with multiple peaks with paired
# relationship and isotope
index2B3 <-
(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
(unique(list$rtcluster) %in% unique(resultiso$rtg)) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B3 <- unique(list$rtcluster)[index2B3]
message(paste(
c(
sum(index2B3),
'group(s) with paired relationship and isotope',
rtg2B3
),
collapse = ' '
))
# print(rtg2B3)
if (sum(index2B3) > 0) {
for (i in seq_along(rtg2B3)) {
# filter the isotope peaks
dfiso <-
resultiso[resultiso$rtg == rtg2B3[i],]
dfpaired <-
resultdiff[resultdiff$rtg == rtg2B3[i],]
if (nrow(dfiso) > 0 & nrow(dfpaired) > 0) {
# remove peaks with more than one isotopes
massstd <-
apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
massstd <-
unique(massstd[!(massstd %in% massstdmax)])
dis <-
stats::dist(massstd, method = "manhattan")
df <- data.frame(
ms1 = massstd[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = massstd[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = round(as.numeric(dis),
digits)
)
# remove the adducts
if (sum((df$diff %in% dfpaired$diff2)) > 0) {
massstd <- unique(apply(df[df$diff %in%
dfpaired$diff2,], 1, function(x)
min(x[1],
x[2])))
massused <-
unique(c(df$ms1, df$ms2))
massadd <-
unique(c(df$ms1[df$diff %in%
dfpaired$diff2], df$ms2[df$diff %in%
dfpaired$diff2]))
massextra <-
massused[!(massused %in% massadd)]
mass <-
c(massextra, massstd)
} else {
mass <- massstd
}
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else if (nrow(dfiso) > 0) {
# remove peaks with more than one peaks
massstd <-
apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
mass <-
unique(massstd[!(massstd %in% massstdmax)])
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else if (nrow(dfpaired) > 0) {
mass <- apply(dfpaired, 1, function(x)
min(x[1],
x[2]))
mass <- unique(mass)
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfpaired$rt,
rtg = dfpaired$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else{
resultstd2B3 <- NULL
}
}
}
# Combine the peaks from rt groups with single ion
resultstd <- rbind(resultstd1,
resultstd2A,
resultstd2B3,
resultstd2B2,
resultstd2B1)
resultstd <- unique(resultstd)
colnames(resultstd) <- c("mz", "rt", "rtg")
resultstd <- as.data.frame(resultstd)
# return the data
list$stdmassindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(round(resultstd$mz, accuracy), resultstd$rtg)
list$stdmass <- resultstd
# use correlation to refine peaks within the same retention groups
if (!is.null(corcutoff)) {
mzo <- NULL
for (i in seq_along(unique(list$rtcluster))) {
resulttemp <- list$data[list$rtcluster == i & list$stdmassindex,]
mz <-
list$mz[list$rtcluster == i &
list$stdmassindex]
cor2 <- stats::cor(t(resulttemp))
df <- data.frame(
ms1 = mz[which(lower.tri(cor2),
arr.ind = TRUE)[, 1]],
ms2 = mz[which(lower.tri(cor2),
arr.ind = TRUE)[, 2]],
cor = cor2[lower.tri(cor2)]
)
df2 <-
apply(df, 1, function(x)
ifelse(abs(x[3]) >= corcutoff, x[1], NA))
df2 <- unique(stats::na.omit(df2))
mz2 <-
paste0(round(mz[(round(mz, accuracy) %in% round(df2, accuracy))], accuracy), '@', i)
mzo <- c(mzo, mz2)
}
list$stdmassindex <-
list$stdmassindex &
(!(paste0(
round(list$mz, accuracy), '@', list$rtcluster
) %in% mzo))
# show message about std mass
n <- sum(list$stdmassindex)
message(paste(n, "std mass found."))
} else{
# show message about std mass
n <- nrow(resultstd)
message(paste(n, "std mass found."))
}
return(list)
}
#' GlobalStd algorithm with structure/reaction directed analysis
#' @param list a peaks list with mass to charge, retention time and intensity data
#' @param rtcutoff cutoff of the distances in cluster, default 10
#' @param ng cutoff of global PMD's retention time group numbers, If ng = NULL, 20 percent of RT cluster will be used as ng, default NULL.
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @param freqcutoff pmd frequency cutoff for structures or reactions, default NULL. This cutoff will be found by PMD network analysis when it is NULL.
#' @param mdrange mass defect range to ignore. Default NULL and c(0.25,0.9) to retain the possible reaction related paired mass
#' @param sda logical, option to perform structure/reaction directed analysis, default FALSE.
#' @return list with GlobalStd algorithm processed data.
#' @examples
#' data(spmeinvivo)
#' re <- globalstd(spmeinvivo)
#' @seealso \code{\link{getpaired}},\code{\link{getstd}},\code{\link{getsda}},\code{\link{plotstd}},\code{\link{plotstdsda}},\code{\link{plotstdrt}}
#' @export
globalstd <- function(list,
rtcutoff = 10,
ng = NULL,
corcutoff = NULL,
digits = 2,
accuracy = 4,
freqcutoff = NULL,
mdrange = NULL,
sda = FALSE) {
list <-
getpaired(
list,
rtcutoff = rtcutoff,
ng = ng,
digits = digits,
accuracy = accuracy,
mdrange = mdrange
)
if (sum(list$pairedindex) > 0) {
list2 <-
getstd(
list,
corcutoff = corcutoff,
digits = digits,
accuracy = accuracy
)
if(sda){
list3 <-
getsda(
list2,
rtcutoff = rtcutoff,
corcutoff = corcutoff,
digits = digits,
freqcutoff = freqcutoff
)
}
} else{
if(sda){
message('no paired relationship, directly go to structure directed analysis.')
list3 <-
getsda(
list,
rtcutoff = rtcutoff,
corcutoff = corcutoff,
digits = digits,
freqcutoff = freqcutoff
)
}else{
message('no paired relationship found.')
}
}
if(sda){
return(list3)
}else{
return(list2)
}
}
#' Get Pseudo-Spectrum as peaks cluster based on correlation analysis.
#' @param list a list with peaks intensity
#' @param corcutoff cutoff of the correlation coefficient, default 0.9
#' @param rtcutoff cutoff of the distances in cluster, default 10
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with Pseudo-Spectrum index
#' @examples
#' data(spmeinvivo)
#' cluster <- getcorcluster(spmeinvivo)
#' @export
getcorcluster <- function(list,
corcutoff = 0.9,
rtcutoff = 10,
accuracy = 4) {
mz <- list$mz
rt <- list$rt
dis <- stats::dist(list$rt, method = "manhattan")
fit <- stats::hclust(dis)
rtcluster <- stats::cutree(fit, h = rtcutoff)
n <- length(unique(rtcluster))
message(paste(n, "retention time cluster found."))
cluster <- mzs <- mzo <- NULL
data <- list$data
for (i in seq_along(unique(rtcluster))) {
# find the mass within RT
if(sum(rtcluster==i)>1){
bin <- data[rtcluster == i,]
if (is.matrix(bin)) {
msdata <- apply(bin, 1, mean)
} else {
msdata <- mean(bin)
}
mzt <- round(mz[rtcluster == i],digits = accuracy)
cor2 <- stats::cor(t(bin))
df <- data.frame(ms1 = mzt[which(lower.tri(cor2),
arr.ind = TRUE)[, 1]],
ms2 = mzt[which(lower.tri(cor2),
arr.ind = TRUE)[, 2]],
cor = cor2[lower.tri(cor2)])
dfc <- df[df$cor>=corcutoff,]
# select larger ions
df2 <- apply(df,1,function(x) ifelse(abs(x[3]) >= corcutoff, x[1], NA))
df2 <- unique(stats::na.omit(df2))
if(nrow(dfc)>0){
mztn <- unique(c(dfc$ms1,dfc$ms2))
nt <- igraph::graph_from_data_frame(dfc)
x <- igraph::components(nt)$membership
ins <-
msdata[match(as.numeric(igraph::V(nt)$name),mzt)]
mztnl <- stats::aggregate(ins,by=list(x),max)
mzc <- mzt[ins %in% mztnl$x]
mzi <- mzt[!(mzt %in% mztn)]
clustert <- NULL
for(j in seq_along(unique(x))){
mzic <- round(as.numeric(igraph::V(nt)$name), accuracy)[x==j]
inst <- ins[x==j]
tdf <- cbind.data.frame(mz=mzic, i=j, rtgt=i, ins=inst)
clustert <- rbind(clustert, tdf)
}
if(length(mzi)>0){
clusterp <- cbind.data.frame(mz=mzi, i=(length(unique(x))+1):(length(unique(x))+length(mzi)), rtgt=i, ins=msdata[match(mzi,mzt)])
cluster <- rbind(cluster, clustert,clusterp)
}else{
cluster <- rbind(cluster, clustert)
}
mzo <- c(mzo, paste0(df2, '@', i))
mzs <- c(mzs, mzc,mzi)
}else{
clustert <- cbind.data.frame(mz=mzt, i=seq_along(mzt), rtgt=i, ins=msdata)
cluster <- rbind(cluster,clustert)
mzo <- c(mzo, paste0(mzt, '@', i))
mzs <- c(mzs, mzt)
}
}else{
mzt <- round(mz[rtcluster == i],digits = accuracy)
clustert <- cbind.data.frame(mz=mzt, i=seq_along(mz[rtcluster==i]), rtgt=i, ins=mean(data[rtcluster==i,]))
cluster <- rbind(cluster,clustert)
mzo <- c(mzo, paste0(mzt, '@', i))
mzs <- c(mzs, mzt)
}
}
list$stdmassindex <-
!(paste0(round(list$mz, accuracy), '@', rtcluster) %in% mzo)
if (!is.null(mzs)) {
list$stdmassindex2 <-
round(list$mz, accuracy) %in% round(mzs, accuracy)
}
cluster$largei <- paste0(cluster$i,'@',cluster$rtgt)
cluster <- cluster[match(round(mz,accuracy),cluster$mz),]
list$rtcluster <- rtcluster
list$cluster <- cluster
return(list)
}
#' Get Pseudo-Spectrum as peaks cluster based on pmd analysis.
#' @param list a list from getstd function
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with Pseudo-Spectrum index
#' @examples
#' data(spmeinvivo)
#' re <- getpaired(spmeinvivo)
#' re <- getstd(re)
#' cluster <- getcluster(re)
#' @seealso \code{\link{getpaired}},\code{\link{getstd}},\code{\link{plotstd}}
#' @export
getcluster <- function(list,
corcutoff = NULL,
accuracy = 4) {
mz <- list$mz[list$stdmassindex]
rt <- list$rt[list$stdmassindex]
rtg <- list$rtcluster[list$stdmassindex]
if (is.null(list$data)) {
message('You need intensity data to use corcutoff and export pseudospectra')
msdata <- NULL
} else{
data <- list$data
if (is.matrix(data)|is.data.frame(data)) {
msdata <- apply(data, 1, sum)
} else {
msdata <-sum(data)
}
}
stdg <- rep('stdgroup', length(list$rtcluster))
# filter high freq ions and find std mass
resultdiff <- list$paired
resultiso <- list$iso
resultmulti <- list$multi
if (!is.null(corcutoff)) {
resultdiff <- resultdiff[abs(resultdiff$cor) >= corcutoff,]
resultiso <- resultiso[resultiso$cor >= corcutoff,]
resultmulti <-
resultmulti[resultmulti$cor >= corcutoff, ]
}
# multi-charger
index1 <-
paste0(round(resultmulti$ms1, accuracy), '@', resultmulti$rtg)
index2 <-
paste0(round(resultmulti$ms2, accuracy), '@', resultmulti$rtg)
# isotope
index3 <-
paste0(round(resultiso$ms1, accuracy), '@', resultiso$rtg)
index4 <-
paste0(round(resultiso$ms2, accuracy), '@', resultiso$rtg)
# diff
index5 <-
paste0(round(resultdiff$ms1, accuracy), '@', resultdiff$rtg)
index6 <-
paste0(round(resultdiff$ms2, accuracy), '@', resultdiff$rtg)
index00 <-
paste0(round(list$mz, accuracy), '@', list$rtcluster)
msdata <- msdata[!duplicated(index00)]
index000 <- unique(index00)
cluster <- mzs <- NULL
for (i in 1:sum(list$stdmassindex)) {
mzt <- mz[i]
rtgt <- rtg[i]
indexstd <- paste0(round(mzt, accuracy), '@', rtgt)
multiover <-
unique(c(resultmulti$ms2[index1 %in% indexstd], resultmulti$ms1[index2 %in% indexstd]))
isoover <-
unique(c(resultiso$ms2[index3 %in% indexstd], resultiso$ms1[index4 %in% indexstd]))
diffover <-
unique(c(resultdiff$ms2[index5 %in% indexstd], resultdiff$ms1[index6 %in% indexstd]))
stdmassg <- c(mzt, multiover, isoover, diffover)
mzx <- list$mz[list$mz %in% stdmassg]
if (!is.null(msdata)) {
index <- paste0(round(mzx, accuracy), '@', rtgt)
ins <- msdata[index000 %in% unique(index)]
tdf <-
cbind.data.frame(mz = mzx[!duplicated(index)], i, rtgt, ins)
} else{
tdf <- cbind.data.frame(stdmassg, i, rtgt)
}
cluster <- rbind.data.frame(cluster, tdf)
}
mergegroup <- function(temp){
t <- any(table(temp$mz)>1)
if(t){
cluster2 <- temp[!duplicated(temp$mz),]
ti<-igraph::components(igraph::graph_from_data_frame(temp))$membership[seq_along(cluster2$mz)]
cluster2$largei <- paste0(ti,'@',cluster2$rtgt)
return(cluster2)
}else{
temp$largei <- paste0(temp$i,'@',temp$rtgt)
return(temp)
}
}
cl <- split(cluster,cluster$rtgt)
cluster <- lapply(cl, mergegroup)
list$cluster <- do.call("rbind", cluster)
if (!is.null(msdata)) {
for(i in unique(list$cluster$largei)){
t <- list$cluster[list$cluster$largei==i,]
mzst <- paste0(unique(round(t$mz[which.max(t$ins)],accuracy)),'@',t$rtgt[1])
mzs <- c(mzs, mzst)
}
}
if (!is.null(mzs)) {
list$stdmassindex2 <-
paste0(round(list$mz, accuracy), '@', list$rtcluster) %in% mzs
}
return(list)
}
#' Compare matrices using PCA similarity factor
#'
#' @param x Matrix with sample in column and features in row
#' @param y Matrix is compared to x.
#' @param dim number of retained dimensions in the comparison. Defaults to all.
#' @return Ratio of projected variance to total variance
#' @references Singhal, A. and Seborg, D. E. (2005), Clustering multivariate time-series data. J. Chemometrics, 19: 427-438. doi: 10.1002/cem.945
#' @author Edgar Zanella Alvarenga
#' @export
#' @examples
#' c1 <- matrix(rnorm(16),nrow=4)
#' c2 <- matrix(rnorm(16),nrow=4)
#' pcasf(c1, c2)
#'
pcasf <- function(x, y, dim = NULL) {
cov.x <- stats::cov(x)
cov.y <- stats::cov(y)
if (is.null(dim)){
dim <- dim(cov.x)[1]
}
eg.x <- eigen(cov.x)
eg.y <- eigen(cov.y)
eg.x.values <- eg.x$values[1:dim]
eg.y.values <- eg.y$values[1:dim]
eg.x.vectors <- eg.x$vectors[,1:dim]
eg.y.vectors <- eg.y$vectors[,1:dim]
total_var <- eg.x.values %*% eg.y.values
return (c(pcasf = sum((eg.x.values %o% eg.y.values) * ((t(eg.x.vectors) %*% (eg.y.vectors))**2))/total_var))
}
yufree/pmd documentation built on April 19, 2024, 9:31 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 pcasf getcluster getcorcluster globalstd getstd getpaired
Documented in getcluster getcorcluster getpaired getstd globalstd pcasf
#' Filter ions/peaks based on retention time hierarchical clustering, paired mass distances(PMD) and PMD frequency analysis.
#' @param list a list with mzrt profile
#' @param rtcutoff cutoff of the distances in retention time hierarchical clustering analysis, default 10
#' @param ng cutoff of global PMD's retention time group numbers, If ng = NULL, 20 percent of RT cluster will be used as ng, default NULL.
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @param mdrange mass defect range to ignore. Default NULL and c(0.25,0.9) to retain the possible reaction related paired mass
#' @return list with tentative isotope, multi-chargers, adducts, and neutral loss peaks' index, retention time clusters.
#' @examples
#' data(spmeinvivo)
#' pmd <- getpaired(spmeinvivo)
#' @seealso \code{\link{getstd}},\code{\link{getsda}},\code{\link{plotpaired}}
#' @export
getpaired <-
function(list,
rtcutoff = 10,
ng = NULL,
digits = 2,
accuracy = 4,
mdrange = NULL) {
# Convert a pmd data.frame to isotope, multi, and diff pmd data.frame list.
rtpmd <- function(bin, i) {
medianrtxi <- stats::median(bin$rt)
if (ncol(bin) > 2) {
if (nrow(bin) > 1) {
# get mz diff
dis <-
stats::dist(bin$mz, method = "manhattan")
# get intensity cor
cor <-
stats::cor(t(bin[, -c(1, 2)]))
df <-
data.frame(
ms1 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = as.numeric(dis),
rt = medianrtxi,
rtg = i,
cor = cor[lower.tri(cor)]
)
# remove isotope
isoindex <-
(round(df$diff, digits) != 0) &
((
df$diff %% 1 < 0.01 &
df$diff >= 1 &
df$diff < 2
) | (
df$diff %% 2 < 0.01 &
df$diff >= 2 &
df$diff < 3
) | (
df$diff %% 1 > 0.99 &
df$diff >= 1 &
df$diff < 2
) | (
df$diff %% 1 > 0.99 &
df$diff >= 0 &
df$diff < 1
)
)
# higher is C13
# massstd <-
# apply(df[isoindex,], 1, function(x)
# min(x[1], x[2]))
massstdmax <-
apply(df[isoindex,], 1, function(x)
max(x[1], x[2]))
isomass <-
unique(massstdmax)
dfiso <- df[isoindex, ]
if (nrow(dfiso) > 0) {
df <-
df[!(df[, 1] %in% isomass) &
!(df[, 2] %in% isomass), ]
}
# remove multi chargers
multiindex <-
(round(df$diff %% 1, 1) == 0.5)
dfmulti <- df[multiindex, ]
mass <-
unique(df[multiindex, 1], df[multiindex, 2])
# From HMDB lowest mass with 0.5 is 394.4539, remove those ions related pmd
multimass <-
mass[round(mass %% 1, 1) == 0.5 & mass<350]
if (nrow(dfmulti) > 0) {
df <-
df[!(df[, 1] %in% multimass) & !(df[, 2] %in% multimass),]
}
dfdiff <- df
return(
list(
dfmulti = dfmulti,
dfiso = dfiso,
dfdiff = dfdiff,
solo = NULL
)
)
} else {
solo <- cbind(bin[, c(1:2)],
rtg = i,
cor = 1)
return(
list(
dfmulti = NULL,
dfiso = NULL,
dfdiff = NULL,
solo = solo
)
)
}
} else{
if (nrow(bin) > 1) {
# get mz diff
dis <-
stats::dist(bin$mz, method = "manhattan")
df <-
data.frame(
ms1 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = bin$mz[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = as.numeric(dis),
rt = medianrtxi,
rtg = i
)
# higher is C13
# massstd <-
# apply(df[isoindex,], 1, function(x)
# min(x[1], x[2]))
massstdmax <-
apply(df[isoindex,], 1, function(x)
max(x[1], x[2]))
isomass <-
unique(massstdmax)
dfiso <- df[isoindex, ]
if (nrow(dfiso) > 0) {
df <-
df[!(df[, 1] %in% isomass) &
!(df[, 2] %in% isomass), ]
}
# remove multi chargers
multiindex <-
(round(df$diff %% 1, 1) == 0.5)
dfmulti <- df[multiindex, ]
mass <-
unique(df[multiindex, 1], df[multiindex, 2])
# From HMDB lowest mass with 0.5 is 394.4539, remove those ions related pmd
multimass <-
mass[round(mass %% 1, 1) == 0.5 & mass<350]
if (nrow(dfmulti) > 0) {
df <-
df[!(df[, 1] %in% multimass) & !(df[, 2] %in% multimass),]
}
dfdiff <- df
return(
list(
dfmulti = dfmulti,
dfiso = dfiso,
dfdiff = dfdiff,
solo = NULL
)
)
} else {
solo <- cbind(bin, rtg = i)
return(
list(
dfmulti = NULL,
dfiso = NULL,
dfdiff = NULL,
solo = solo
)
)
}
}
}
# paired mass diff analysis
dis <- stats::dist(list$rt, method = "manhattan")
fit <- stats::hclust(dis)
rtcluster <- stats::cutree(fit, h = rtcutoff)
n <- length(unique(rtcluster))
message(paste(n, "retention time cluster found."))
# automate ng selection when ng is NULL
ng <- ifelse(is.null(ng), round(n * 0.2), ng)
if (!is.null(list$data)) {
groups <- cbind.data.frame(mz = list$mz,
rt = list$rt,
list$data)
} else {
groups <- cbind.data.frame(mz = list$mz,
rt = list$rt)
message('No intensity data!')
}
split <- split.data.frame(groups, rtcluster)
rtpmdtemp <-
mapply(rtpmd,
split,
as.numeric(names(split)),
SIMPLIFY = FALSE)
result <- do.call(Map, c(rbind, rtpmdtemp))
# filter the list get the rt cluster
list$rtcluster <- rtcluster
result$dfdiff$diff2 <-
round(result$dfdiff$diff, digits)
result$dfdiff$md <- result$dfdiff$diff%%1
# speed up by mass defect
if(is.null(mdrange)){
list$diff <- result$dfdiff
}else{
list$diff <- result$dfdiff <- result$dfdiff[result$dfdiff$md<mdrange[1]|result$dfdiff$md>mdrange[2],]
}
# speed up by group
pmd <-
as.numeric(names(table(result$dfdiff$diff2)[table(result$dfdiff$diff2) > ng]))
# pmd <- unique(result$dfdiff$diff2)
idx <- NULL
for (i in seq_along(pmd)) {
l <-
length(unique(result$dfdiff[result$dfdiff$diff2 == pmd[i], 'rtg']))
idx <-
c(idx, ifelse(l > ng, TRUE, FALSE))
}
pmd2 <- pmd[idx]
list$paired <-
result$dfdiff[result$dfdiff$diff2 %in% pmd2, ]
if (nrow(result$dfdiff) > 0) {
list$pairedindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(list$paired$ms1, accuracy),
round(list$paired$ms2,
accuracy)
),
c(list$paired$rtg, list$paired$rtg))
}
# get the data index by rt groups with high frequencies
# PMD
list$diffindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfdiff$ms1, accuracy),
round(result$dfdiff$ms2,
accuracy)
),
c(result$dfdiff$rtg, result$dfdiff$rtg))
# get the data index by rt groups with single ions
if (!is.null(result$solo)) {
list$soloindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(round(result$solo$mz, accuracy),
result$solo$rtg)
list$solo <- result$solo
}
# get the data index by rt groups with isotope ions
if (!is.null(result$dfiso)) {
list$isoindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfiso$ms1, accuracy),
round(result$dfiso$ms2, accuracy)
),
c(result$dfiso$rtg, result$dfiso$rtg))
list$iso <- result$dfiso
}
# get the data index by rt groups with multi charger ions
if (!is.null(result$dfmulti)) {
list$multiindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(c(
round(result$dfmulti$ms1, accuracy),
round(result$dfmulti$ms2,
accuracy)
),
c(result$dfmulti$rtg, result$dfmulti$rtg))
list$multi <- result$dfmulti
}
# show message about std mass
message(paste(sum(list$pairedindex), "paired masses found "))
message(
paste(
length(unique(list$paired$diff2)),
"unique within RT clusters high frequency PMD(s) used for further investigation."
)
)
message(paste(c("The unique within RT clusters high frequency PMD(s) is(are) ", unique(list$paired$diff2)), collapse=" "), '.')
message(paste(
sum(list$isoindex),
"isotopologue(s) related paired mass found."
))
message(paste(
sum(list$multiindex),
"multi-charger(s) related paired mass found."
))
# return results
return(list)
}
#' Find the independent ions for each retention time hierarchical clustering based on PMD relationship within each retention time cluster and isotope and return the index of the std data for each retention time cluster.
#' @param list a list from getpaired function
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with std mass index
#' @examples
#' data(spmeinvivo)
#' pmd <- getpaired(spmeinvivo)
#' std <- getstd(pmd)
#' @seealso \code{\link{getpaired}},\code{\link{getsda}},\code{\link{plotstd}}
#' @export
getstd <-
function(list,
corcutoff = NULL,
digits = 2,
accuracy = 4) {
resultstd2A <- resultstd2B1 <- resultstd2B2 <- resultstd2B3 <- NULL
# filter high freq ions and find std mass
resultdiff <- list$paired
resultiso <- list$iso
if (!is.null(corcutoff)) {
resultdiff <- resultdiff[abs(resultdiff$cor) >= corcutoff,]
resultiso <- resultiso[resultiso$cor >= corcutoff,]
}
# filter the mass from mass pairs within retention time
# group group 1: RT groups with solo peak
resultstd1 <- NULL
if (!is.null(list$solo)) {
resultstd1 <- cbind(list$solo$mz, list$solo$rt,
list$solo$rtg)
}
message(paste(
c(
sum(list$soloindex),
"retention group(s) have single peaks.",
list$rtcluster[list$soloindex]
),
collapse = ' '
))
# group 2: RT groups with multiple peaks group 2A: RT
# groups with multiple peaks while no isotope/paired
# relationship
index2A <-
!(
unique(list$rtcluster) %in% unique(resultdiff$rtg) |
unique(list$rtcluster) %in% unique(resultiso$rtg)
) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2A <- unique(list$rtcluster)[index2A]
message(paste(
c(
sum(index2A),
'group(s) with multiple peaks while no isotope/paired relationship',
rtg2A
),
collapse = " "
))
# print(rtg2A)
if (sum(index2A) > 0) {
for (i in seq_along(rtg2A)) {
mass <- list$mz[list$rtcluster == rtg2A[i]]
rt <- list$rt[list$rtcluster == rtg2A[i]]
mass <- max(mass)
suppressWarnings(resultstdtemp <-
c(mass, stats::median(rt),
rtg2A[i]))
suppressWarnings(resultstd2A <-
rbind(resultstd2A,
resultstdtemp))
}
}
# group 2B: RT groups with multiple peaks with
# isotope/paired relationship index2B <-
# (unique(list$rtcluster) %in%
# unique(resultdiff$rtg))|(unique(list$rtcluster) %in%
# unique(resultiso$rtg)) group 2B1: RT groups with
# multiple peaks with isotope without paired
# relationship
index2B1 <-
!(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
unique(list$rtcluster) %in% unique(resultiso$rtg) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B1 <- unique(list$rtcluster)[index2B1]
# print(rtg2B1)
message(paste(
c(
sum(index2B1),
'group(s) with multiple peaks with isotope without paired relationship',
rtg2B1
),
collapse = " "
))
if (sum(index2B1) > 0) {
for (i in seq_along(rtg2B1)) {
# filter the isotope peaks
dfiso <-
resultiso[resultiso$rtg == rtg2B1[i],]
if (nrow(dfiso) > 0) {
massstd <- apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
mass <-
unique(massstd[!(massstd %in% massstdmax)])
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
resultstd2B1 <-
rbind(resultstd2B1, resultstdtemp)
} else{
resultstd2B1 <- NULL
}
}
}
# group 2B2: RT groups with multiple peaks with paired
# relationship without isotope
index2B2 <-
(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
!(unique(list$rtcluster) %in% unique(resultiso$rtg)) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B2 <- unique(list$rtcluster)[index2B2]
message(paste(
c(
sum(index2B2),
'group(s) with paired relationship without isotope',
rtg2B2
),
collapse = ' '
))
# print(rtg2B2)
if (sum(index2B2) > 0) {
for (i in seq_along(rtg2B2)) {
# filter the paired peaks
df <-
resultdiff[resultdiff$rtg == rtg2B2[i],]
if (nrow(df) > 0) {
mass <- apply(df, 1, function(x)
min(x[1],
x[2]))
mass <- unique(mass)
suppressWarnings(
resultstdtemp <-
cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = df$rt,
rtg = df$rtg
)
)
suppressWarnings(
resultstd2B2 <- rbind(resultstd2B2,
resultstdtemp)
)
}
}
}
# group 2B3: RT groups with multiple peaks with paired
# relationship and isotope
index2B3 <-
(unique(list$rtcluster) %in% unique(resultdiff$rtg)) &
(unique(list$rtcluster) %in% unique(resultiso$rtg)) &
!(unique(list$rtcluster) %in% unique(list$solo$rtg))
rtg2B3 <- unique(list$rtcluster)[index2B3]
message(paste(
c(
sum(index2B3),
'group(s) with paired relationship and isotope',
rtg2B3
),
collapse = ' '
))
# print(rtg2B3)
if (sum(index2B3) > 0) {
for (i in seq_along(rtg2B3)) {
# filter the isotope peaks
dfiso <-
resultiso[resultiso$rtg == rtg2B3[i],]
dfpaired <-
resultdiff[resultdiff$rtg == rtg2B3[i],]
if (nrow(dfiso) > 0 & nrow(dfpaired) > 0) {
# remove peaks with more than one isotopes
massstd <-
apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
massstd <-
unique(massstd[!(massstd %in% massstdmax)])
dis <-
stats::dist(massstd, method = "manhattan")
df <- data.frame(
ms1 = massstd[which(lower.tri(dis),
arr.ind = TRUE)[, 1]],
ms2 = massstd[which(lower.tri(dis),
arr.ind = TRUE)[, 2]],
diff = round(as.numeric(dis),
digits)
)
# remove the adducts
if (sum((df$diff %in% dfpaired$diff2)) > 0) {
massstd <- unique(apply(df[df$diff %in%
dfpaired$diff2,], 1, function(x)
min(x[1],
x[2])))
massused <-
unique(c(df$ms1, df$ms2))
massadd <-
unique(c(df$ms1[df$diff %in%
dfpaired$diff2], df$ms2[df$diff %in%
dfpaired$diff2]))
massextra <-
massused[!(massused %in% massadd)]
mass <-
c(massextra, massstd)
} else {
mass <- massstd
}
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else if (nrow(dfiso) > 0) {
# remove peaks with more than one peaks
massstd <-
apply(dfiso, 1, function(x)
min(x[1],
x[2]))
massstdmax <-
apply(dfiso, 1, function(x)
max(x[1],
x[2]))
mass <-
unique(massstd[!(massstd %in% massstdmax)])
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfiso$rt,
rtg = dfiso$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else if (nrow(dfpaired) > 0) {
mass <- apply(dfpaired, 1, function(x)
min(x[1],
x[2]))
mass <- unique(mass)
suppressWarnings(
resultstdtemp <- cbind(
mz = ifelse(is.na(mass), NULL, c(mass)),
rt = dfpaired$rt,
rtg = dfpaired$rtg
)
)
suppressWarnings(
resultstd2B3 <- rbind(resultstd2B3,
resultstdtemp)
)
} else{
resultstd2B3 <- NULL
}
}
}
# Combine the peaks from rt groups with single ion
resultstd <- rbind(resultstd1,
resultstd2A,
resultstd2B3,
resultstd2B2,
resultstd2B1)
resultstd <- unique(resultstd)
colnames(resultstd) <- c("mz", "rt", "rtg")
resultstd <- as.data.frame(resultstd)
# return the data
list$stdmassindex <-
paste(round(list$mz, accuracy), list$rtcluster) %in%
paste(round(resultstd$mz, accuracy), resultstd$rtg)
list$stdmass <- resultstd
# use correlation to refine peaks within the same retention groups
if (!is.null(corcutoff)) {
mzo <- NULL
for (i in seq_along(unique(list$rtcluster))) {
resulttemp <- list$data[list$rtcluster == i & list$stdmassindex,]
mz <-
list$mz[list$rtcluster == i &
list$stdmassindex]
cor2 <- stats::cor(t(resulttemp))
df <- data.frame(
ms1 = mz[which(lower.tri(cor2),
arr.ind = TRUE)[, 1]],
ms2 = mz[which(lower.tri(cor2),
arr.ind = TRUE)[, 2]],
cor = cor2[lower.tri(cor2)]
)
df2 <-
apply(df, 1, function(x)
ifelse(abs(x[3]) >= corcutoff, x[1], NA))
df2 <- unique(stats::na.omit(df2))
mz2 <-
paste0(round(mz[(round(mz, accuracy) %in% round(df2, accuracy))], accuracy), '@', i)
mzo <- c(mzo, mz2)
}
list$stdmassindex <-
list$stdmassindex &
(!(paste0(
round(list$mz, accuracy), '@', list$rtcluster
) %in% mzo))
# show message about std mass
n <- sum(list$stdmassindex)
message(paste(n, "std mass found."))
} else{
# show message about std mass
n <- nrow(resultstd)
message(paste(n, "std mass found."))
}
return(list)
}
#' GlobalStd algorithm with structure/reaction directed analysis
#' @param list a peaks list with mass to charge, retention time and intensity data
#' @param rtcutoff cutoff of the distances in cluster, default 10
#' @param ng cutoff of global PMD's retention time group numbers, If ng = NULL, 20 percent of RT cluster will be used as ng, default NULL.
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param digits mass or mass to charge ratio accuracy for pmd, default 2
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @param freqcutoff pmd frequency cutoff for structures or reactions, default NULL. This cutoff will be found by PMD network analysis when it is NULL.
#' @param mdrange mass defect range to ignore. Default NULL and c(0.25,0.9) to retain the possible reaction related paired mass
#' @param sda logical, option to perform structure/reaction directed analysis, default FALSE.
#' @return list with GlobalStd algorithm processed data.
#' @examples
#' data(spmeinvivo)
#' re <- globalstd(spmeinvivo)
#' @seealso \code{\link{getpaired}},\code{\link{getstd}},\code{\link{getsda}},\code{\link{plotstd}},\code{\link{plotstdsda}},\code{\link{plotstdrt}}
#' @export
globalstd <- function(list,
rtcutoff = 10,
ng = NULL,
corcutoff = NULL,
digits = 2,
accuracy = 4,
freqcutoff = NULL,
mdrange = NULL,
sda = FALSE) {
list <-
getpaired(
list,
rtcutoff = rtcutoff,
ng = ng,
digits = digits,
accuracy = accuracy,
mdrange = mdrange
)
if (sum(list$pairedindex) > 0) {
list2 <-
getstd(
list,
corcutoff = corcutoff,
digits = digits,
accuracy = accuracy
)
if(sda){
list3 <-
getsda(
list2,
rtcutoff = rtcutoff,
corcutoff = corcutoff,
digits = digits,
freqcutoff = freqcutoff
)
}
} else{
if(sda){
message('no paired relationship, directly go to structure directed analysis.')
list3 <-
getsda(
list,
rtcutoff = rtcutoff,
corcutoff = corcutoff,
digits = digits,
freqcutoff = freqcutoff
)
}else{
message('no paired relationship found.')
}
}
if(sda){
return(list3)
}else{
return(list2)
}
}
#' Get Pseudo-Spectrum as peaks cluster based on correlation analysis.
#' @param list a list with peaks intensity
#' @param corcutoff cutoff of the correlation coefficient, default 0.9
#' @param rtcutoff cutoff of the distances in cluster, default 10
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with Pseudo-Spectrum index
#' @examples
#' data(spmeinvivo)
#' cluster <- getcorcluster(spmeinvivo)
#' @export
getcorcluster <- function(list,
corcutoff = 0.9,
rtcutoff = 10,
accuracy = 4) {
mz <- list$mz
rt <- list$rt
dis <- stats::dist(list$rt, method = "manhattan")
fit <- stats::hclust(dis)
rtcluster <- stats::cutree(fit, h = rtcutoff)
n <- length(unique(rtcluster))
message(paste(n, "retention time cluster found."))
cluster <- mzs <- mzo <- NULL
data <- list$data
for (i in seq_along(unique(rtcluster))) {
# find the mass within RT
if(sum(rtcluster==i)>1){
bin <- data[rtcluster == i,]
if (is.matrix(bin)) {
msdata <- apply(bin, 1, mean)
} else {
msdata <- mean(bin)
}
mzt <- round(mz[rtcluster == i],digits = accuracy)
cor2 <- stats::cor(t(bin))
df <- data.frame(ms1 = mzt[which(lower.tri(cor2),
arr.ind = TRUE)[, 1]],
ms2 = mzt[which(lower.tri(cor2),
arr.ind = TRUE)[, 2]],
cor = cor2[lower.tri(cor2)])
dfc <- df[df$cor>=corcutoff,]
# select larger ions
df2 <- apply(df,1,function(x) ifelse(abs(x[3]) >= corcutoff, x[1], NA))
df2 <- unique(stats::na.omit(df2))
if(nrow(dfc)>0){
mztn <- unique(c(dfc$ms1,dfc$ms2))
nt <- igraph::graph_from_data_frame(dfc)
x <- igraph::components(nt)$membership
ins <-
msdata[match(as.numeric(igraph::V(nt)$name),mzt)]
mztnl <- stats::aggregate(ins,by=list(x),max)
mzc <- mzt[ins %in% mztnl$x]
mzi <- mzt[!(mzt %in% mztn)]
clustert <- NULL
for(j in seq_along(unique(x))){
mzic <- round(as.numeric(igraph::V(nt)$name), accuracy)[x==j]
inst <- ins[x==j]
tdf <- cbind.data.frame(mz=mzic, i=j, rtgt=i, ins=inst)
clustert <- rbind(clustert, tdf)
}
if(length(mzi)>0){
clusterp <- cbind.data.frame(mz=mzi, i=(length(unique(x))+1):(length(unique(x))+length(mzi)), rtgt=i, ins=msdata[match(mzi,mzt)])
cluster <- rbind(cluster, clustert,clusterp)
}else{
cluster <- rbind(cluster, clustert)
}
mzo <- c(mzo, paste0(df2, '@', i))
mzs <- c(mzs, mzc,mzi)
}else{
clustert <- cbind.data.frame(mz=mzt, i=seq_along(mzt), rtgt=i, ins=msdata)
cluster <- rbind(cluster,clustert)
mzo <- c(mzo, paste0(mzt, '@', i))
mzs <- c(mzs, mzt)
}
}else{
mzt <- round(mz[rtcluster == i],digits = accuracy)
clustert <- cbind.data.frame(mz=mzt, i=seq_along(mz[rtcluster==i]), rtgt=i, ins=mean(data[rtcluster==i,]))
cluster <- rbind(cluster,clustert)
mzo <- c(mzo, paste0(mzt, '@', i))
mzs <- c(mzs, mzt)
}
}
list$stdmassindex <-
!(paste0(round(list$mz, accuracy), '@', rtcluster) %in% mzo)
if (!is.null(mzs)) {
list$stdmassindex2 <-
round(list$mz, accuracy) %in% round(mzs, accuracy)
}
cluster$largei <- paste0(cluster$i,'@',cluster$rtgt)
cluster <- cluster[match(round(mz,accuracy),cluster$mz),]
list$rtcluster <- rtcluster
list$cluster <- cluster
return(list)
}
#' Get Pseudo-Spectrum as peaks cluster based on pmd analysis.
#' @param list a list from getstd function
#' @param corcutoff cutoff of the correlation coefficient, default NULL
#' @param accuracy measured mass or mass to charge ratio in digits, default 4
#' @return list with Pseudo-Spectrum index
#' @examples
#' data(spmeinvivo)
#' re <- getpaired(spmeinvivo)
#' re <- getstd(re)
#' cluster <- getcluster(re)
#' @seealso \code{\link{getpaired}},\code{\link{getstd}},\code{\link{plotstd}}
#' @export
getcluster <- function(list,
corcutoff = NULL,
accuracy = 4) {
mz <- list$mz[list$stdmassindex]
rt <- list$rt[list$stdmassindex]
rtg <- list$rtcluster[list$stdmassindex]
if (is.null(list$data)) {
message('You need intensity data to use corcutoff and export pseudospectra')
msdata <- NULL
} else{
data <- list$data
if (is.matrix(data)|is.data.frame(data)) {
msdata <- apply(data, 1, sum)
} else {
msdata <-sum(data)
}
}
stdg <- rep('stdgroup', length(list$rtcluster))
# filter high freq ions and find std mass
resultdiff <- list$paired
resultiso <- list$iso
resultmulti <- list$multi
if (!is.null(corcutoff)) {
resultdiff <- resultdiff[abs(resultdiff$cor) >= corcutoff,]
resultiso <- resultiso[resultiso$cor >= corcutoff,]
resultmulti <-
resultmulti[resultmulti$cor >= corcutoff, ]
}
# multi-charger
index1 <-
paste0(round(resultmulti$ms1, accuracy), '@', resultmulti$rtg)
index2 <-
paste0(round(resultmulti$ms2, accuracy), '@', resultmulti$rtg)
# isotope
index3 <-
paste0(round(resultiso$ms1, accuracy), '@', resultiso$rtg)
index4 <-
paste0(round(resultiso$ms2, accuracy), '@', resultiso$rtg)
# diff
index5 <-
paste0(round(resultdiff$ms1, accuracy), '@', resultdiff$rtg)
index6 <-
paste0(round(resultdiff$ms2, accuracy), '@', resultdiff$rtg)
index00 <-
paste0(round(list$mz, accuracy), '@', list$rtcluster)
msdata <- msdata[!duplicated(index00)]
index000 <- unique(index00)
cluster <- mzs <- NULL
for (i in 1:sum(list$stdmassindex)) {
mzt <- mz[i]
rtgt <- rtg[i]
indexstd <- paste0(round(mzt, accuracy), '@', rtgt)
multiover <-
unique(c(resultmulti$ms2[index1 %in% indexstd], resultmulti$ms1[index2 %in% indexstd]))
isoover <-
unique(c(resultiso$ms2[index3 %in% indexstd], resultiso$ms1[index4 %in% indexstd]))
diffover <-
unique(c(resultdiff$ms2[index5 %in% indexstd], resultdiff$ms1[index6 %in% indexstd]))
stdmassg <- c(mzt, multiover, isoover, diffover)
mzx <- list$mz[list$mz %in% stdmassg]
if (!is.null(msdata)) {
index <- paste0(round(mzx, accuracy), '@', rtgt)
ins <- msdata[index000 %in% unique(index)]
tdf <-
cbind.data.frame(mz = mzx[!duplicated(index)], i, rtgt, ins)
} else{
tdf <- cbind.data.frame(stdmassg, i, rtgt)
}
cluster <- rbind.data.frame(cluster, tdf)
}
mergegroup <- function(temp){
t <- any(table(temp$mz)>1)
if(t){
cluster2 <- temp[!duplicated(temp$mz),]
ti<-igraph::components(igraph::graph_from_data_frame(temp))$membership[seq_along(cluster2$mz)]
cluster2$largei <- paste0(ti,'@',cluster2$rtgt)
return(cluster2)
}else{
temp$largei <- paste0(temp$i,'@',temp$rtgt)
return(temp)
}
}
cl <- split(cluster,cluster$rtgt)
cluster <- lapply(cl, mergegroup)
list$cluster <- do.call("rbind", cluster)
if (!is.null(msdata)) {
for(i in unique(list$cluster$largei)){
t <- list$cluster[list$cluster$largei==i,]
mzst <- paste0(unique(round(t$mz[which.max(t$ins)],accuracy)),'@',t$rtgt[1])
mzs <- c(mzs, mzst)
}
}
if (!is.null(mzs)) {
list$stdmassindex2 <-
paste0(round(list$mz, accuracy), '@', list$rtcluster) %in% mzs
}
return(list)
}
#' Compare matrices using PCA similarity factor
#'
#' @param x Matrix with sample in column and features in row
#' @param y Matrix is compared to x.
#' @param dim number of retained dimensions in the comparison. Defaults to all.
#' @return Ratio of projected variance to total variance
#' @references Singhal, A. and Seborg, D. E. (2005), Clustering multivariate time-series data. J. Chemometrics, 19: 427-438. doi: 10.1002/cem.945
#' @author Edgar Zanella Alvarenga
#' @export
#' @examples
#' c1 <- matrix(rnorm(16),nrow=4)
#' c2 <- matrix(rnorm(16),nrow=4)
#' pcasf(c1, c2)
#'
pcasf <- function(x, y, dim = NULL) {
cov.x <- stats::cov(x)
cov.y <- stats::cov(y)
if (is.null(dim)){
dim <- dim(cov.x)[1]
}
eg.x <- eigen(cov.x)
eg.y <- eigen(cov.y)
eg.x.values <- eg.x$values[1:dim]
eg.y.values <- eg.y$values[1:dim]
eg.x.vectors <- eg.x$vectors[,1:dim]
eg.y.vectors <- eg.y$vectors[,1:dim]
total_var <- eg.x.values %*% eg.y.values
return (c(pcasf = sum((eg.x.values %o% eg.y.values) * ((t(eg.x.vectors) %*% (eg.y.vectors))**2))/total_var))
}
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.