-
Todor Kondic authoredTodor Kondic authored
Code owners
Assign users and groups as approvers for specific file changes. Learn more.
mix.R 38.67 KiB
## Copyright (C) 2020 by University of Luxembourg
## Licensed under the Apache License, Version 2.0 (the "License");
## you may not use this file except in compliance with the License.
## You may obtain a copy of the License at
## http://www.apache.org/licenses/LICENSE-2.0
## Unless required by applicable law or agreed to in writing, software
## distributed under the License is distributed on an "AS IS" BASIS,
## WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
## See the License for the specific language governing permissions and
## limitations under the License.
stripext<-function(fn) {
bits<-strsplit(fn,split="\\.")[[1]]
if (length(bits)> 1) paste(head(bits,-1),collapse=".") else fn}
get_mz_cmp_l<-function(id,adduct,cmpL) {
ind<-match(id,cmpL$ID)
mz<-cmpL$mz[[ind]]
smiles<-cmpL$SMILES[[ind]]
res<-if (!is.null(mz) && !is.na(mz)) {
mz
} else if (nchar(smiles)>0)
{
mde<-as.character(adduct)
wh<-ADDUCTMAP[[mde]]
RChemMass::getSuspectFormulaMass(smiles)[[wh]]
} else stop("Both SMILES and mz fields, for ID ",id,", found empty in the compound list. Aborting.")
res
}
calc_mz_from_formula_outer <- function(chform,adduct,id) {
check_chform <- enviPat::check_chemform(ISOTOPES,chform)
wind <- which(check_chform$warning)
if (length(wind) > 0) stop("Cannot understand the following formulas: ",
paste(check_chform$new_formula[wind],collapse = ","))
mol_form <- check_chform$new_formula
l_mol <- length(mol_form)
l_add <- length(adduct)
adds <- ADDUCTS[Name %in% adduct,.(Name,
add=as.character(Formula_add),
ded=as.character(Formula_ded),
charge=Charge)]
dt <- dtable(ID = rep(id,each = l_add),
mol_form = rep(mol_form,each = l_add),
adduct = rep(adds$Name,l_mol),
add = rep(adds$add,l_mol),
ded = rep(adds$ded,l_mol),
charge= rep(adds$charge,l_mol))
merger <- function (mol_form,add,ded) {
full_form <- rep(NA_character_,length(mol_form))
both_ind <- which(add != 'FALSE' & ded != 'FALSE')
add_only_ind <- which(add != 'FALSE' & ded == 'FALSE')
ded_only_ind <- which(ded != 'FALSE' & add == 'FALSE')
ainds <- c(both_ind,add_only_ind)
full_form[ainds] <- vapply(ainds,function (i) enviPat::mergeform(mol_form[[i]],add[[i]]),FUN.VALUE = character(1), USE.NAMES = F)
dinds <- c(both_ind,ded_only_ind)
full_form[dinds] <- vapply(dinds,function (i) {
z <- check_ded2(mol_form[[i]],ded[[i]])
if (z) enviPat::subform(mol_form[[i]],ded[[i]]) else NA_character_
},
FUN.VALUE = character(1))
full_form
}
dt[,("full_form"):=.(merger(mol_form,add,ded))] dt[!is.na(full_form),("mz"):=.(mapply(function(ff,ch) enviPat::isopattern(ISOTOPES,chemforms = ff,
charge = ch, verbose = F)[[1]][1],
full_form,
charge, USE.NAMES = F))]
dt[is.na(full_form),("mz"):=NA_real_]
dt
}
calc_mz_from_formula <- function(chform,adduct,id) {
check_chform <- enviPat::check_chemform(ISOTOPES,chform)
wind <- which(check_chform$warning)
if (length(wind) > 0) stop("Cannot understand the following formulas: ",
paste(check_chform$new_formula[wind],collapse = ","))
mol_form <- check_chform$new_formula
uad <- unique(adduct)
uadds <- lapply(uad,function(a) ADDUCTS[Name==a,.(Name,
add=as.character(Formula_add),
ded=as.character(Formula_ded),
charge=Charge),on=""])
names(uadds) <- uad
adds <- rbindlist(l=lapply(adduct,function(a) uadds[[a]]))
merger <- function (mol_form,add,ded) {
res <- numeric(length(mol_form))
both_ind <- which(add != 'FALSE' & ded != 'FALSE')
add_only_ind <- which(add != 'FALSE' & ded == 'FALSE')
ded_only_ind <- which(ded != 'FALSE' & add == 'FALSE')
ainds <- c(both_ind,add_only_ind)
res[ainds] <- vapply(ainds,function (i) enviPat::mergeform(mol_form[[i]],add[[i]]),FUN.VALUE = character(1), USE.NAMES = F)
dinds <- c(both_ind,ded_only_ind)
res[dinds] <- vapply(dinds,function (i) {
z <- check_ded2(mol_form[[i]],ded[[i]])
if (z) enviPat::subform(mol_form[[i]],ded[[i]]) else NA_character_
},
FUN.VALUE = character(1))
res
}
forms <- merger(mol_form,adds$add,adds$ded)
## Check if formulas actually calculated.
bad_idx <- which(forms=="0")
bad_adducts <- adduct[bad_idx]
bad_ids <- id[bad_idx]
non_dups <- !duplicated(bad_idx)
bad_ids <- bad_ids[non_dups]
bad_adducts <- bad_adducts[non_dups]
if (length(bad_idx)>0) stop(paste0("Unable to process the adducts:\n",
paste(bad_adducts,collapse = ","),
"\nfor id-s:",
paste(bad_ids,collapse = ",")))
mz <- the_ifelse(!is.na(forms),
mapply(function(ff,ch) enviPat::isopattern(ISOTOPES,chemforms = ff,
charge = ch, verbose = F)[[1]][1],
forms,
adds$charge, USE.NAMES = F),
NA_real_)
mz
}
calc_mz_from_smiles <- function(smiles,adduct,id) {
mol <- lapply(smiles,function(s) try(RMassBank::getMolecule(s), silent = T))
check <- which(is.atomic(mol))
if (length(check) > 0)
stop("Errors in SMILES with IDs:",paste(id[which],collapse = ','))
mol_form <- sapply(mol,function(x) (rcdk::get.mol2formula(x))@string,USE.NAMES = F)
names(mol_form) <- id
calc_mz_from_formula(mol_form,adduct,id)
}
calc_mz_from_smiles_outer <- function(smiles,adduct,id) {
mol <- lapply(smiles,function(s) try(RMassBank::getMolecule(s), silent = T))
check <- which(is.atomic(mol))
if (length(check) > 0)
stop("Errors in SMILES with IDs:",paste(id[which],collapse = ','))
mol_form <- sapply(mol,function(x) (rcdk::get.mol2formula(x))@string,USE.NAMES = F)
names(mol_form) <- id
calc_mz_from_formula_outer(mol_form,adduct,id)
}
get_col_from_cmp_l<-function(id,cname,cmpL) {
ind<-match(id,cmpL$ID)
x<-cmpL[[cname]][[ind]]
if (!is.null(x)) x else NA
}
gen_clean_state_summ<-function(summ) {
summ$Comments <- ""
summ[c("MS1","MS2","Alignment","AboveNoise")] <- T
summ["MS2rt"] <- NA_real_
summ["iMS2rt"] <- NA_integer_
summ["rt"]<-NA_real_
summ["checked"]<-'NONE'
summ
}
pp_touch_q<-function(summ) {
## Returns indices that are ok to be auto processed.
which(summ$checked==SUMM_CHK_NONE | summ$checked==SUMM_CHK_AUTO)
}
preProc <- function (summ,noiseFac=3,errRT=0.5,intThreshMS1=1e5,intThreshMS2=5000.) {
wds<-unique(summ$wd)
fn_spec<-function(wd) readRDS(file.path(wd,FN_SPEC))
message("Loading RDS-es ...")
allData<-lapply(wds,fn_spec)
names(allData)<-wds
message("... done with RDSs")
## QA check plan:
##
## If MS1 does not exist, set MS1 to F, as well as everything else except MS2.
## If it exists, proceed to noise check.
## If noise check fails, set AboveNoise and Alignment to F.
##
##
## MS2 will be checked independently.
## If MS2 does not exist, set MS2 and Alignment to F.
## If it does, check the Alignment.
## If Alignment is wrong, set Alignment to F.
##
## Terminology: MS1 does not exist if the intensity is below the
## intensity threshold. MS2 does not exist if it was not picked up
## during the dataframe generation stage. In this case, the file
## with the corresponding ID will not be there.
okinds<- pp_touch_q(summ)
for (ind in okinds) {
wd <- summ$wd[ind]
id <- summ$ID[ind]
eics<-allData[[wd]]$eic nid<-id2name(id)
ii<-match(nid,MSnbase::fData(eics)[["ID"]]) #id, because id-s, not nid-s are in fData for ms1 eics;
eic1<-eics[[ii]]
eic<-data.frame(rt=MSnbase::rtime(eic1)/60.,intensity=MSnbase::intensity(eic1))
colnames(eic)<-c("rt","intensity")
maxInt <- NULL
if (nrow(eic)==0) {
warning("No chromatogram for id ",id," found in", wd, " . Skipping.")
next
}
ms1MaxInd<-which.max(eic$intensity)
maxInt<-eic$intensity[[ms1MaxInd]]
summ[ind,"rt"]<-eic$rt[[ms1MaxInd]]
##If MS1 does not exist, set entry to F.
if (maxInt < intThreshMS1) {
summ[ind,"MS1"] <- F
## Other checks automatically fail, too.
summ[ind,"Alignment"] <- F
summ[ind,"AboveNoise"] <- F
} else {
## Noisy?
if (summ[ind,"AboveNoise"]) {
mInt <- mean(eic$intensity)
if (maxInt < noiseFac*mInt) {
summ[ind,"AboveNoise"] <- F
summ[ind,"Alignment"] <- F ## If noisy, this is
## probably meaningles, so
## F.
}
}
}
## MS2 checks.
ms2<-allData[[wd]]$ms2
ms2nids<-names(ms2)
if (! (nid %in% ms2nids)) {
summ[ind,"MS2"] <- F
summ[ind,"Alignment"] <- F
} else {
sp<-ms2[[nid]]
## Alignment still makes sense to be checked?
if (summ[ind,"Alignment"]) {
## rtInd <- ms1MaxInd #match(maxInt,eic$intensity)
rtMS1Peak <- eic$rt[[ms1MaxInd]]
msms<-MSnbase::fData(sp)[,c("rtm","maxI")]
colnames(msms)<-c("rt","intensity")
rtInd <- which((msms$rt > rtMS1Peak - errRT) &
(msms$rt < rtMS1Peak + errRT)) #Close enough?
rtIndMS1 <- which((eic$rt > rtMS1Peak - errRT) &
(eic$rt < rtMS1Peak + errRT)) #Filter the relevant MS1 part.
eicFilt<- eic[rtIndMS1,]
eicFilt<- eicFilt[which(eicFilt$intensity>intThreshMS1),]
mInt<- maxInt #mean(eicFilt$intensity)
rtInd <- rtInd[which(msms$intensity[rtInd]>intThreshMS2)] #Intense enough?
msmsRT <- msms$rt[rtInd]
msmsInt<- msms$intensity[rtInd]
if (length(msmsRT) > 0) {
msmsRTind <- which.min(abs(msmsRT - rtMS1Peak))
summ[ind,"iMS2rt"] <- rtInd[msmsRTind]
summ[ind,"MS2rt"] <- msmsRT[msmsRTind]
} else {
summ[ind,"Alignment"] <- F
} }
}
summ[ind,"checked"]<-SUMM_CHK_AUTO
}
summ
}
smiles2img <- function(smiles, kekulise=TRUE, width=300, height=300,
zoom=1.3,style="cow", annotate="off", abbr="on",suppressh=TRUE,
showTitle=FALSE, smaLimit=100, sma=NULL) {
dep <- rcdk::get.depictor(width = width, height = height, zoom = zoom, style = style, annotate = annotate,
abbr = abbr, suppressh = suppressh, showTitle = showTitle, smaLimit = smaLimit,
sma = NULL)
mol <- RMassBank::getMolecule(smiles)
z<-rcdk::view.image.2d(mol, depictor=dep)
grid::rasterGrob(z)
}
gen_ms2_spec_data <- function(id,tag,iMS2rt,data,luckyN=NA) {
## Given the id, tag and the index of the MS2 spectrum, return the
## dataframe of the spectrum, with luckyN number of lagerst
## intensity peaks.
nid<-id2name(id)
if (!is.na(iMS2rt)) {
d <- data$ms2[[nid]][[iMS2rt]]
if (!is.null(d)) {
x<-data.frame(mz=MSnbase::mz(d),intensity=MSnbase::intensity(d))
res<-if (!is.na(luckyN)) {
ord<-order(x$intensity,decreasing = T)
len<-length(ord)
sz<-min(len,luckyN)
nx<-x[ord,]
nx<-nx[1:sz,]
ord<-order(nx$mz)
nx[ord,]
} else x
return(res)
} else {
return(NULL)}
} else return(NULL)
}
gen_ms2_spec_fn <- function(id,tag,adduct,set,width=6) {
suppressWarnings({
iid<-as.numeric(id)
iid<- if (!is.na(iid)) iid else id
num <- formatC(iid,width = width,format='d',flag='0')
ss<-trimws(paste(num,adduct,tag,set,sep="_"),which='both')
paste(ss,".csv",sep='')
})
}
plot_id_msn <- function(ni,
data,
rtMS1,
rtMS2,
iMS2rt,
mass,
smile,
tags, summ,
prop,
theme,
pal="Dark2",
cex=0.75,
rt_digits=2,
m_digits=4) {
clean_range<-function(def,rng) {
x1 <- rng[1]
x2 <- rng[2]
if (is.na(x1) || x1 == 0) x1 <- def[1]
if (is.na(x2) || x2 == 0) x2 <- def[2]
c(x1,x2)
}
mk_title<-function() paste("EIC (",
"m/z = ",
formatC(mass,format='f',digits=m_digits),
")",sep='')
mk_leg_lab<-function(tag,rt) {paste(tag,"; rt= ",formatC(rt[[tag]],format='f',digits=rt_digits),"min")}
sci10<-function(x) {ifelse(x==0, "0", parse(text=gsub("[+]", "", gsub("e", " %*% 10^", scales::scientific_format()(x)))))}
i<-name2id(ni)
dfChrMS1<-NULL
dfChrMS2<-NULL
dfSpecMS2<-NULL
## MS1 time series.
dfschrms1<-lapply(tags,function(tag) {d<-data[[tag]]$eic
ind<-match(ni,MSnbase::fData(d)[["ID"]])
cg<-d[[ind]]
data.frame(rt=MSnbase::rtime(cg)/60.,
intensity=MSnbase::intensity(cg),tag=as.character(tag),legend=mk_leg_lab(tag,rtMS1))
})
dfChrMS1<-do.call(rbind,c(dfschrms1,list(make.row.names=F)))
## MS2 spectral time series.
dfsChrMS2<-lapply(tags,function(tag) {
d<-data[[tag]]$ms2[[ni]]
if (!is.null(d)) {
df<-MSnbase::fData(d)[,c("rtm","maxI")]
colnames(df)<-c("rt","intensity")
df$tag<-as.character(tag)
df$legend=mk_leg_lab(tag,rtMS2)
df
} else NULL
})
dfsChrMS2<-dfsChrMS2[!is.null(dfsChrMS2)]
if (!all(sapply(dfsChrMS2,is.null))) dfChrMS2<-do.call(rbind,c(dfsChrMS2,list(make.row.names=F)))
## MS2 Spectrum.
if (!all(sapply(dfsChrMS2,is.null))) {
dfsSpecMS2<-lapply(tags,function(tag) {
d<-data[[tag]]$ms2[[ni]]
if (!is.null(d)) {
ind<-iMS2rt[[tag]]
if (!is.na(ind)) {
x<-data.frame(mz=MSnbase::mz(d[[ind]]),intensity=MSnbase::intensity(d[[ind]]))
x$tag<-tag x
} else NULL
}
})
dfsSpecMS2<-dfsSpecMS2[!is.null(dfsSpecMS2)]
dfSpecMS2<-do.call(rbind,c(dfsSpecMS2,list(make.row.names=F)))
}
## Ranges
if (!is.null(dfChrMS1)) {
rrtMS1<-range(dfChrMS1$rt)
rrtMS1 <- if (is.null(prop$ms1$rt)) rrtMS1 else clean_range(rrtMS1,prop$ms1$rt)
rrtMS2<-rrtMS1
rintMS1<-range(dfChrMS1$intensity)
rintMS1 <- if (is.null(prop$ms1$irng)) rintMS2 else clean_range(rintMS1,prop$ms1$irng)
}
if (!is.null(dfChrMS2)) {
rrtMS2 <- if (is.null(prop$ms2$rt)) rrtMS2 else clean_range(rrtMS2,prop$ms2$rt)
rintMS2<-range(dfChrMS2$intensity)
rintMS2 <- if (is.null(prop$ms2$irng)) rintMS2 else clean_range(rintMS2,prop$ms2$irng)
}
if (is.data.frame(dfSpecMS2)) {
rmzSpMS2<-range(dfSpecMS2$mz)
rintSpMS2<-range(dfSpecMS2$intensity)
rmzSpMS2<- if (is.null(prop$spec$mzrng)) rmzSpMS2 else clean_range(rmzSpMS2,prop$spec$mzrng)
rintSpMS2<- if (is.null(prop$spec$irng)) rintSpMS2 else clean_range(rintSpMS2,prop$spec$irng)
}
ch_ms1_deco<-function(ggobj) {
titMS1<-mk_title()
scale_y<-if (!prop$ms1$axis=="log") {
ggplot2::scale_y_continuous
} else {
ggplot2::scale_y_log10
}
ggobj+
ggplot2::geom_line(ggplot2::aes(colour=legend),key_glyph=KEY_GLYPH)+
ggplot2::coord_cartesian(xlim = rrtMS1,
ylim = rintMS1)+
ggplot2::labs(x=CHR_GRAM_X,y=CHR_GRAM_Y,
title=titMS1,tag=i,
colour=PLOT_MS1_LEG_TIT)+
scale_y(labels=sci10)+theme()
}
ch_ms2_deco<-function(ggobj) {
scale_y<-if (!prop$ms2$axis=="log") {
ggplot2::scale_y_continuous
} else {
ggplot2::scale_y_log10
}
ggobj+
ggplot2::geom_linerange(ggplot2::aes(colour=legend),key_glyph=KEY_GLYPH)+
ggplot2::coord_cartesian(xlim = rrtMS2,
ylim = rintMS2)+
ggplot2::labs(x=CHR_GRAM_X,y=CHR_GRAM_Y,title=NULL,subtitle = "MS2",tag = " ")+
scale_y(labels=sci10)+
ggplot2::labs(colour=PLOT_MS2_LEG_TIT)+theme()
}
ch_spec_deco<-function(ggobj) {
scale_y<-if (!prop$spec$axis=="log") { ggplot2::scale_y_continuous
} else {
ggplot2::scale_y_log10
}
ggobj+
ggplot2::geom_linerange(ggplot2::aes(colour=tag),key_glyph=KEY_GLYPH)+
ggplot2::coord_cartesian(xlim = rmzSpMS2,
ylim = rintSpMS2)+
ggplot2::labs(subtitle="MS2",y="intensity")+
scale_y(labels=sci10)+theme()
}
## MS1 time series.
plMS1<- if(is.data.frame(dfChrMS1) && nrow(dfChrMS1)>0) {
ch_ms1_deco(ggplot2::ggplot(data=dfChrMS1,ggplot2::aes(x=rt,y=intensity,group=legend)))
} else NULL
## Empty
plEmpty<-ggplot2::ggplot(data=dfChrMS1,ggplot2::aes(x=rt,y=intensity))+ggplot2::theme_void()
## MS2 time series.
plMS2 <- if (!all(sapply(dfsChrMS2,is.null))) {
ch_ms2_deco(ggplot2::ggplot(data=dfChrMS2,ggplot2::aes(x=rt,ymin=0,ymax=intensity,group=legend)))
} else plEmpty
## Structure
if (!is.null(smile) && !is.na(smile) && !nchar(smile)<1) {
g<-smiles2img(smile,width=500,height=500,zoom=4.5)
plStruc<-ggplot2::ggplot(data=dfChrMS1,ggplot2::aes(x=rt,y=intensity))+
ggplot2::geom_blank()+ggplot2::annotation_custom(g)+ggplot2::theme_void()
} else plStruc<-plEmpty
## MS2 Spectrum
if (!all(sapply(dfsChrMS2,is.null))) {
plSpecMS2<-if (is.data.frame(dfSpecMS2)) { #sometimes
#dfSpecMS2 ends up
#as a list of
#logicals; this
#probably happens
#when either MS2 is
#bad in some way,
#or the RT
#intervals are
#mismatched.
ch_spec_deco(ggplot2::ggplot(data=dfSpecMS2,
ggplot2::aes(x=mz,
ymin=0,
ymax=intensity,group=tag)))
} else plEmpty
} else plSpecMS2<-plEmpty
## Lucky N the most intense N TODO
## lckN<-if (is.data.frame(dfSpecMS2)) {
## ord<-order(dfSpecMS2$intensity,decreasing=T)
## ll<-length(ord)
## theL<-min(ll,MS2_1ST_N)
## mzN<-dfSpecMS2$mz[ord][1:theL]
## inN<-dfSpecMS2$intensity[ord][1:theL]
## df<-data.frame("m/z"=mzN,"intensity"=inN) ## message("DF:")
## str(df)
## message("---DF")
## gridExtra::tableGrob(df) #+ggplot2::labs(subtitle="Top m/z")
## } else NULL
res<- if (!is.null(plMS1)) cowplot::plot_grid(plMS1,plStruc,plMS2,plEmpty,plSpecMS2,align = "hv",axis='l',ncol = 2,nrow=3,rel_widths=c(3,1)) else NULL
res
}
add_wd_to_mzml <- function(fn,proj) {
wd<-basename(tools::file_path_sans_ext(fn))
file.path(proj,wd)
}
getEntryFromComp<-function(entry,id,set,adduct,compTab) {
ind <- which(compTab$ID %in% id &
compTab$set %in% set &
compTab$adduct %in% adduct)
res<- if (length(ind)==1) compTab[ind,entry] else {
if (length(ind)>1) {
warning("Nonunique selection in comprehensive table:")
for (i in ind) {
message('ID: ',compTab$ID[[i]],' set: ',compTab$set[[i]],' adduct: ',compTab$adduct[[i]])
}
warning("The compound set table likely containes duplicate IDs per set/adduct combination. Please correct this.")
} else {
warning("Entries not found for id ", id,"set ",set, "and adduct ", adduct, " .")
}
}
res
names(res)<-entry
res
}
## add_comp_summ <- function(ft,ctab) {
## nR<-nrow(ft)
## mzCol<-rep(NA,nR)
## nmCol<-rep("",nR)
## rtCol<-rep(NA,nR)
## for (ir in 1:nR) {
## id<-ft[ir,"ID"]
## set<-ft[ir,"set"]
## m<-ft[ir,"adduct"]
## entries<-getEntryFromComp(c("mz","Name","rt"),id,set,m,ctab)
## mzCol[[ir]]<- entries[["mz"]]
## nm<-entries[["Name"]]
## nmCol[[ir]]<- if (!is.na(nm)) nm else ""
## rtCol[[ir]]<- entries[["rt"]]
## }
## ft$mz<-mzCol
## ft$Name<-nmCol
## ft$rt<-rtCol
## ft
## }
get_set_adduct <- function(s,mzml) {
unique(mzml[set == s,adduct])
}
vald_comp_tab<-function(df,ndf,checkSMILES=F,checkMz=F,checkNames=F) {
## Fields.
if (is.null(df$ID)) stop("Column ID missing in ",ndf," .")
if (checkMz && is.null(df$mz)) stop("Column mz missing in ", ndf, " .")
if (checkSMILES && is.null(df$SMILES)) stop("Column SMILES missing in", ndf, " .")
if (checkNames && is.null(df$Name)) warning("Column Name missing in ", ndf," , continuing without.")
if (is.null(df$RT) && is.null(df$rt)) {
warning("Column RT (alternatively, rt) missing in ", ndf, ", continuing without.")
} else {
if (is.null(df$rt)) {
df$rt<-df$RT
df$RT<-NULL
}
}
## Missing IDs?
ind<-which(is.na(df$ID))
if (length(ind)>0) {
for (i in ind) {
warning("ID missing at row: ",i," . Big trouble ahead.")
}
return(NULL)
}
## Unique IDs?
luids<-length(unique(df$ID))
if (length(df$ID) > luids) {
warning("Duplicate IDs in ", ndf, " are not allowed.")
return(NULL)
}
## Missing SMILES?
if (checkSMILES) {
ind<-which(is.na(df$SMILES))
if (length(ind)>0) {
for (i in ind) {
warning("SMILES missing at row: ",i, "; ID: ",df$ID[[i]]," .")
}
return(NULL)
}
lsmiles<-nrow(df)
ll<-length(unique(df$SMILES))
if (ll<lsmiles) {
warning("There are duplicate SMILES in the compound list. Trouble ahead.")
}
}
## Missing mz?
if (checkMz) {
ind<-which(is.na(df$mz))
if (length(ind)>0) {
for (i in ind) {
warning("mz missing at row: ",i, "; ID: ",df$ID[[i]]," .")
}
return(NULL)
}
}
df
}
read_setid <- function(fn,cmpds) {
assert(file.exists(fn),msg=paste("Please provide valid compounds set table:", fn))
assert(nrow(cmpds) > 0,msg="Please provide at least one compounds list.")
setid <- file2tab(fn,colClasses=c(ID="character"))
x<-cmpds[setid,on='ID'][,.SD,.SDcols=c(colnames(setid),'known')]
sids <- unique(setid$ID)
cids <- unique(cmpds$ID)
diff <- setdiff(sids,cids)
assert(length(diff)==0,msg=paste("The following IDs from set table have not been found in the compound table:","------",print_table(dtable(diff)),"------",sep = "\n"))
x
}
write_conf <- function(m,fn) {
m$conf$data <- get_fn_ftab(m)
yaml::write_yaml(x=m$conf,file=fn)
}
write_state <- function(m,fn_conf) {
write_conf(m,fn_conf)
tab2file(tab=m$input$tab$mzml,file=file.path(m$conf$project,FN_DATA_TAB))
}
read_conf <- function(fn) {
cf <- yaml::yaml.load_file(fn)
fnl <- cf$compound$lists
if (length(fnl)>0) {
nms <- character(0)
for (i in 1:length(fnl)) {
nms <- gen_uniq_lab(nms,pref = 'L')
}
names(fnl) <- nms
}
cf$compound$lists <- fnl
## conf_trans(cf)
cf
}
##' @export
get_fn_comp <- function(m) {
file.path(m$conf$project,FN_COMP_TAB)
}
##' @export
get_fn_summ <- function(m) {
file.path(m$conf$project, FN_SUMM)
}
##' @export
get_fn_extr <- function(m) {
file.path(m$conf$project, "extracted.rds")
}
##' @export
get_fn_conf <- function(m) {
file.path(m$conf$project, FN_CONF)
}
##' @export
get_fn_ftab <- function(m) {
file.path(m$conf$project, FN_DATA_TAB)
}
init_state <- function(m) {
m$out$tab <- list()
m$input$datafiles <- NULL
m$input$tab$mzml <- EMPTY_MZML
lab <- gen_uniq_lab(list(),pref="L")
m$input$tab$lists <- list()
m$input$tab[[lab[[1]]]] <- EMPTY_CMPD_LIST
m
}
base_conf <- function () {
m <- list()
m$conf <- list(project=getwd(), compounds=list(lists=list(),
sets="",
data=""),
extr=list(fn=""),
debug = F)
m
}
extr_conf <- function(m) {
m$conf$tolerance <- list("ms1 coarse"=MS1_ERR_COARSE,
"ms1 fine"=MS1_ERR_FINE,
"eic"=EIC_ERR,
"rt"=RT_EXTR_ERR)
m
}
presc_conf <- function(m) {
m$conf$prescreen <- list("ms1_int_thresh"=1e5,
"ms2_int_thresh"=2.5e3,
"s2n"=3,
"ret_time_shift_tol"=0.5)
m
}
new_conf <- function() presc_conf(
extr_conf(
base_conf()))
verify_cmpd_l <- function(dt,fn) {
fields <- colnames(EMPTY_CMPD_LIST)
dtflds <- colnames(dt)
assert('ID' %in% dtflds, msg = paste('ID column must be present and filled in', fn))
ess <- c('SMILES','Formula','mz')
pres <- ess %in% dtflds
assert(length(pres) > 0,
msg = paste('Compound list from ',fn,
'does not contain any of "SMILES", "Formula", or "mz". \nThe compound list needs at least one of those to be valid.'))
exst <- ess[pres]
x <- lapply(exst,function (nm) do.call(all,as.list(is.na(dt[[nm]]))))
assert(!do.call(all,x), msg = paste('At least one of', paste(exst,collapse = ','),
'\nmust contain some values in compound list from',fn))
invisible(T)
}
## INPUT TRANSLATORS
#' @export
grab_unit <- function(entry,unit) {
what <- paste0("\\<",unit,"\\>$")
entry <- trimws(entry,which="both")
if (grepl(what,entry))
suppressWarnings(as.numeric(sub(paste0("^(.*)",unit),"\\1",entry))) else NA_real_
}
rt_in_min <- function(entry) {
xs <- grab_unit(entry,"s")
xm <- grab_unit(entry,"min")
x <- if (is.na(xm)) xs/60. else xm
x
}
conf_trans_pres <- function(pres_list) {
## Translate and validate prescreening input. pres_list[CONF_PRES_NUM] <- sapply(pres_list[CONF_PRES_NUM],as.numeric)
for (par in CONF_PRES_NUM) {
assert(!suppressWarnings(is.na(pres_list[[par]])),msg=paste("Prescreen parameter",par,"is not a number."))
}
for (par in CONF_PRES_TU) {
xs <- grab_unit(pres_list[[par]],"s")
xm <- grab_unit(pres_list[[par]],"min")
x <- if (is.na(xm)) xs else xm
assert(!is.na(x),msg = paste("Time unit parameter error for",par,"Only s(econds) or min(utes) allowed."))
pres_list[[par]] <- x
}
pres_list
}
## PRESCREENING
create_qa_table <- function(ms,conf_presc) {
## The first input argument is the extracted `ms`, table
## containing MS1 and MS2 spectra. The argument `conf_presc` is
## m$conf$prescreen, the prescreening parameters given in the conf
## file.
## The qa table is just a copy of ms with added quality control
## columns QA_COLS.
## The QA_FLAGS columns are flags specifying which properties of
## compounds are known well, or not.
## For each compound (mass) we ask the following questions:
## qa_ms1_exists -- does the MS1 spectrum exist at all?
## qa_ms2_exists -- do we have any MS2 spectra at all?
## qa_ms1_above_noise -- is MS1 above the noise treshold?
## qa_ms2_near -- is there any MS2 spectrum inside the tolerated
## retention time window around the MS1 peak? That is, are we
## non-RT-shifted?
## qa_ms2_good_int -- Is there any MS2 spectral intensity greater
## than the MS2 threshold and less than the MS1 peak?
## qa_pass -- did the spectrum pass all the checks?
## The columns in QA_NUM_REAL are:
##
## ms1_int -- the maximum intensity of MS1 spectrum over the
## entire run;
##
## ms1_rt -- the retention time of the peak MS1.
## The columns in QA_NUM_INT are:
##
## ms2_sel -- index of the selected MS2 spectrum; if not NA, the
## associated spectrum passed all the checks (qa_pass == T); the
## spectrum itself is in one of the member sublists of the `spec'
## column. The integer `ms2_sel' is then the index of the spectrum
## in that sublist.
##
## ms1_rt_ind -- TODO (but not important to end users).
qa <- list(prescreen=conf_presc)
qa$ms <- data.table::copy(ms)
qa$ms[,(QA_FLAGS):=T] # All checks true by default. Dangerous,
# but we need to believe in our
# filters. Also, the humans who check the
# results. :)
qa$ms[,(QA_NUM_INT):=NA_integer_]
qa$ms[,(QA_NUM_REAL):=NA_real_]
qa
}
assess_ms1 <- function(m) { qa <- m$qa
## Calculate auxiliary variables and indices.
qa$ms[,c("ms1_rt_ind"):=.(sapply(eicMS1,function(e) which.max(e$intensity)))]
qa$ms[length(ms1_rt_ind)==0,("ms1_rt_ind"):=NA_integer_]
qa$ms[,c("ms1_rt","ms1_int","ms1_mean"):=.(NA_real_,NA_real_,NA_real_)]
qa$ms[!is.na(ms1_rt_ind),c("ms1_int","ms1_rt","ms1_mean"):=.(mapply(function (e,i) e$intensity[[i]],eicMS1,ms1_rt_ind),
mapply(function (e,i) e$rt[[i]],eicMS1,ms1_rt_ind),
mapply(function (e,i) mean(e$intensity),eicMS1,ms1_rt_ind))]
check_ms1 <- function(qa) {
qa$ms[(!is.na(ms1_int)),"qa_ms1_exists" := .(ms1_int > qa$prescreen$ms1_int_thresh)]
qa$ms[is.na(ms1_int),("qa_ms1_exists"):=F]
qa$ms[(!qa_ms1_exists),(QA_FLAGS):=F]
qa
}
check_ms1_noise <- function(qa) {
qa$ms[(qa_ms1_exists==T),"qa_ms1_above_noise" := .(ms1_int > qa$prescreen$s2n*ms1_mean)]
qa$ms[(!qa_ms1_above_noise),c("qa_ms2_good_int","qa_ms2_near","qa_ms2_exists","qa_pass"):=F]
qa
}
qa <- check_ms1_noise(check_ms1(qa))
m$qa <- qa
m
}
assess_ms2 <- function(m) {
presconf <- conf_trans_pres(m$conf$prescreen)
## This function takes a spectral list, looks for the members
## inside the retention time window and returns either the indices
## of those that are, or NA.
pick_ms2_rtwin <- function(rtMS1,sp_list,rt_win) {
rt <- sapply(sp_list,function (x) x$rt)
rtl <- rtMS1 - rt_win/2.
rtr <- rtMS1 + rt_win/2.
which(rt > rtl & rt < rtr)
}
## Only return the index which satisfies the intensity
## range.
pick_ms2_int <- function(sp_list,int_lo,int_hi) {
ints <- sapply(sp_list,function (x) max(x$spec$intensity))
which(int_lo < ints & ints < int_hi)
}
## Test only rows that passed MS1 checks and have MS2 spec. To
## test existence of MS2, it is only necessary to make sure that
## either spec member sublist has more than one entry, or if not,
## that the single entry in the sublist is not NA.
m$qa$ms[qa_ms1_exists==T,qa_ms2_exists := .(sapply(spec,function (sl) length(sl)>1 || !is.na(sl[[1]])))]
irows <- which(m$qa$ms$qa_ms1_exists & m$qa$ms$qa_ms2_exists)
rt_win <- 2 * presconf$ret_time_shift_tol
## List of lists of spec indices where MS2 are within the rt
## window.
okind_rt_ms2 <- m$qa$ms[irows, ][, .(tmp=mapply(function (rt1,spl) pick_ms2_rtwin(rt1,spl,rt_win),
ms1_rt,
spec,
USE.NAMES=F,
SIMPLIFY=F))]$tmp m$qa$ms[irows,"qa_ms2_near"] <- sapply(okind_rt_ms2,function (x) length(x) > 0)
m$qa$ms[-irows,"qa_ms2_near"] <- F
## List of lists of spec indices where MS2 are within the desired
## intensity range.
okind_int_ms2 <- m$qa$ms[irows, ][, .(tmp=mapply(pick_ms2_int,
spec,
presconf$ms2_int_thresh,
ms1_int,
SIMPLIFY=F))]$tmp
m$qa$ms[irows,"qa_ms2_good_int"] <- sapply(okind_int_ms2,function (x) length(x) > 0)
m$qa$ms[-irows,"qa_ms2_good_int"] <- F
## Candidates for the MS2 choices.
okind <- mapply(intersect,okind_int_ms2,okind_rt_ms2)
m$qa$ms[irows,"qa_pass"] <- sapply(okind,function (x) length(x) > 0)
m$qa$ms[-irows,"qa_pass"] <- F
## Throw out the possibly empty members.
really_okind <- okind[m$qa$ms[irows, ]$qa_pass]
m$qa$ms[which(qa_pass),ms2_sel:=.(mapply(function (spl,inds,ms1rt) {
rtdiff <- sapply(spl[inds],function (x) abs(x$rt-ms1rt))
closest <- which.min(rtdiff)
inds[[closest]]
},
spec,
really_okind,
ms1_rt,
SIMPLIFY=T))]
m
}
gen_mz_err_f <- function(entry,msg) {
eppm <- grab_unit(entry,"ppm")
eda <- grab_unit(entry,"Da")
shinyscreen:::assert(xor(is.na(eda), is.na(eppm)), msg = msg)
if (is.na(eda)) function(mz) eppm*1e-6*mz else function (mz) eda
}
gen_rt_err <- function(entry,msg) {
em <- grab_unit(entry,"min")
es <- grab_unit(entry,"s")
shinyscreen:::assert(xor(is.na(em), is.na(es)), msg = msg)
if (is.na(em)) es/60. else em
}
fig_path <- function(top,set,group,id,suff,ext="pdf") {
base <- paste("plot",set,group,id,suff,sep="_")
fn <- paste0(base,".",ext)
fn <- gsub("\\[","",fn)
fn <- gsub("\\]","",fn)
fn <- gsub("\\+","p",fn)
fn <- gsub("-","m",fn)
if (!is.null(top)) file.path(top,fn) else fn
}
get_coord_lim <- function(new,def) {
if (is.null(new)) return(def)
res <- new
if (length(new[[1]])==0) res[[1]]<-def[[1]]
if (length(new[[2]])==0) res[[2]]<-def[[2]]
res
}
gen_base_ms1_plot_tab <- function(summ,ms1_spec) {
ident <- c("set",
"adduct",
"tag",
"ID",
"mz",
"Files")
res <- summ[ms1_spec,c(.SD,
list(rt_peak=i.ms1_rt,
eicMS1=lapply(i.eicMS1,list))),
.SDcols=ident,
on=BASE_KEY,
nomatch=NULL]
setkeyv(res,cols=BASE_KEY)
res
}
plot_decor <- function(m,islog,all_labels,legend_name) {
textf <- ggplot2::element_text
## Logarithmic, or linear y axis?
scale_y <- if (shiny::isTruthy(islog))
ggplot2::scale_y_log10 else ggplot2::scale_y_continuous
rt_new_lim <- c(rt_in_min(m$conf$figures$rt_min),
rt_in_min(m$conf$figures$rt_max))
rt_lim <- get_coord_lim(rt_new_lim,DEFAULT_RT_RANGE)
my_theme <- function (...) ggplot2::theme()
getpal <- colorRampPalette(RColorBrewer::brewer.pal(8,"Dark2"))
col_all_vals <- getpal(length(all_labels))
names(col_all_vals) <- all_labels
scale_colour <- function(breaks, labels, ...) ggplot2::scale_colour_manual(values = col_all_vals,
breaks = breaks,
labels = labels,
name = legend_name,...)
my_coord <- ggplot2::coord_cartesian(xlim = rt_lim)
function(plot,breaks,labels) plot + my_coord + scale_colour(breaks=breaks,
labels=labels) + scale_y() + my_theme()
}
plot_eic_ms1 <- function(df,style_fun,plot_label) {
mk_leg_lab<-function(tag,rt) {if (length(tag) > 0) paste(tag,"; rt= ",formatC(rt,format='f',digits=RT_DIGITS)," min",sep='') else character(0)}
mz <- df[,unique(mz)]
tbl <- df[,.(verb_labs=mk_leg_lab(plot_label,rt_peak),plot_label),
by=c("plot_label","rt_peak")]
verb_labs <- tbl[,verb_labs]
labs <- tbl[,plot_label]
df[,plot_label:=factor(plot_label)]
style_fun(ggplot2::ggplot(df,ggplot2::aes(x=rt,y=intensity,colour=plot_label)),
breaks=labs,
labels=verb_labs) +
ggplot2::geom_line(key_glyph=KEY_GLYPH) +
ggplot2::labs(x=CHR_GRAM_X,
y=CHR_GRAM_Y)
}