readme.md

Vincent et al. 2019

EM_mito.R

#install.packages(c("mixOmics","RVAideMemoire"))
library(mixOmics)
library(RVAideMemoire)

# Calculate whether measure is greater in control group after scaling
# Used to colour points in VIP plots
direction=function(dt,measure){
  dts = as.data.frame(scale(dt[,-1]))
  dts$Group = dt$Group
  res = median(dts[[measure]][dts$Group=="Control"],na.rm=TRUE) > median(dts[[measure]][dts$Group!="Control"],na.rm=TRUE)
  return(res)
}

or = rgb(247/255,158/255,84/255) # orange
bl = rgb(69/255,150/255,207/255) # blue

gr1 = rgb(116/255,161/255,77/255) # green
or1 = rgb(247/255,161/255,30/255) # orange

bl2 = rgb(3/255,40/255,210/255) # blue
or2 = rgb(68/255,186/255,82/255) # orange

# Read in data
dat = read.delim("Summary stats Control, patient and mouse 05-03-18.csv",sep=",",stringsAsFactors=FALSE)
rownames(dat) = dat$Subject
dat$Subject = NULL
rownames(dat)

# Transpose and ensure data are numeric
dat = data.frame(t(dat),stringsAsFactors=FALSE)
for(col in colnames(dat)[2:length(colnames(dat))]) dat[[col]]=as.numeric(as.character(dat[[col]]))

# Tidy measure names
colnames(dat) = gsub("MCI.Vol","MCIperVol",colnames(dat))
colnames(dat) = gsub("\\.","\n",colnames(dat))
colnames(dat) = gsub("\nmitos","",colnames(dat))
colnames(dat) = gsub("X\n","Percent",colnames(dat))
colnames(dat) = gsub("\nmito","",colnames(dat))
colnames(dat) = gsub("95\n\nCI","interval",colnames(dat))

# Pairwise correlation plot for examining relationship between measures
colours = c("black","red","blue")
names(colours) = c("Control", "Mito Disease", "Mouse")
pdat = dat[,-1]
pdf("pairwise.pdf",width=30,height=30,pointsize=17.5)
  pairs(pdat,col = colours[dat$Group],upper.panel = NULL, pch = 16, cex = 0.6)
dev.off()

# Consider dropping subset of measures
rejectroot = c("Mean","interval","Kurtosis","MCI","Vol","MCIperVol")
rejectmeasures = unlist(lapply(rejectroot,grep,colnames(dat)))

# Consider keeping subset of variables
keeproot = c("Nanotunnels\nper\n100","Volume\ndensity","Percent\nsimple","Percent\ncomplex","Percent\nsmall","Percent\nlarge","MCI\nMedian","Vol\nMedian")
keepmeasures = c(1,match(keeproot,colnames(dat)))

subsets = c("SUBSET1","SUBSET2","ALL")

# Consider two different comparisons: Control-Patient and Human(Control)-Mouse
keeps = list()
keeps[["Controls & Patients"]] = c("Control","Mito Disease")
keeps[["Controls & Mice"]] = c("Control","Mouse")

def.par = par(no.readonly = TRUE) # save default, for resetting...

for(subset in subsets){

  if(subset=="SUBSET2") datsub=dat[,keepmeasures]
  if(subset=="SUBSET1") datsub=dat[,-rejectmeasures]
  if(subset=="ALL") datsub=dat


  # Supervised learning with PLS-DA: find combinations of measures that best discriminate
  # between two groups.  PLS-DA also ranks original measures by their contribution to 
  # splitting categories (VIP score).  Number of components = 2, for ease of plotting.

  # Multi-page PDF report showing PLS-DA biplots and VIP scores
  pdf(paste("PLSDA",paste(subset,".pdf",sep=""),sep="_"),width=8.27*2.5,height=8.27,pointsize=18)
  layout(matrix(c(1,2,3),nrow=1,ncol=3,byrow=TRUE),widths=c(1,0.5,1))
  for (k in names(keeps)){
    keepcases = keeps[[k]]
    if ("Mouse" %in% keepcases){colcontrol=gr1; colother=or1}else{colcontrol=bl2; colother=or2}
    dt = datsub[datsub$Group%in%keepcases,]
    dt2 = dt[-6,]
    da = plsda(dt2[,-1], factor(dt2$Group), scale = TRUE, ncomp=2)
    op2=par(mar=c(4.5,4.5, 4, 2) + 0.1)
    plotIndiv(da,ellipse=TRUE,style="graphics",title=k,size.axis=1.5,size.xlabel=1.5,size.ylabel=1.5,col=c(colcontrol,colother),cex=1.5)
    par(op2)
    op3 = par(mar=c(4.5,7.0, 4, 2) + 0.1)
    tb = PLSDA.VIP(da)$tab
    if(subset=="ALL") {
       labs = gsub("\n"," ",rownames(tb))
       ulim=2
    }else{
       labs = gsub("\nper\n100","\nper 100",rownames(tb))
       ulim=1.5
    } 
    vcols = ifelse(sapply(rownames(tb),direction,dt = dt),colcontrol,colother)
    plot(tb$VIP,seq(length(tb$VIP),1),xlab="VIP",ylab="",axes=FALSE,type="n",xlim=c(0,ulim), cex.axis=1.5, cex.lab=1.5)
    abline(h = seq(length(tb$VIP),1),col="grey",lwd=2)
    abline(v=1,col="red",lty=2,lwd=2)
    points(tb$VIP,seq(length(tb$VIP),1),pch=15,cex=1.5,col=vcols[rownames(tb)])
    axis(1, cex.axis=1.5, cex.lab=1.5)
    axis(2, labels=labs,at = seq(length(tb$VIP),1),las=2, cex.axis=1.5, cex.lab=1.5)
    #legend("bottomright",legend=c("Higher in controls","Lower in controls"),col=c(colcontrol,colother),pch=15,bg="white")
    par(op3)
    plot(dt[[rownames(tb)[1]]],dt[[rownames(tb)[2]]],
      col=ifelse(dt$Group=="Control",colcontrol,colother),pch=16,cex=2,xlab=gsub("\n"," ",rownames(tb)[1]),ylab=gsub("\n"," ",rownames(tb)[2]),
      cex.axis=1.5, cex.lab=1.5)
    text(dt[[rownames(tb)[1]]],dt[[rownames(tb)[2]]],labels=rownames(dt),col=ifelse(dt$Group=="Control",colcontrol,colother),pos=4,cex=1.5)
    legend("topright",legend=keepcases,col=c(colcontrol,colother),pch=16,cex=1.5)
  }
  dev.off()

  # Unsupervised PCA also splits data nicely, however PCA results generally more difficult to 
  # interpret than PLS-DA for 2-way comparison.  Number of components = 2, for ease of plotting.

  # Multi-page PDF report showing PCA biplots and loading vectors
  pdf(paste("PCA",paste(subset,".pdf",sep=""),sep="_"),width=8.27*2,height=8.27)
  op = par(mfrow=c(1,2),mar=c(4,2, 4, 2) + 0.1)
  for (k in names(keeps)){
    keepcases = keeps[[k]]
    dt = datsub[datsub$Group%in%keepcases,]
    prc = pca(dt[,-1], scale = TRUE, center = TRUE, ncomp=2)
    biplot(prc,main=k,xlim=c(-0.6,0.6))
  }
  par(op)
  dev.off()
}

Licence.txt

MIT License

Copyright (c) 2018 Conor Lawless

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.