ericminikel · November 18, 2013 19:48
diff --git a/antiprion-vs-all.r b/antiprion-vs-all.r
 # Eric Minikel
 # CureFFI.org
 # 2013-11-06
 # Anti-prion compounds plotted in the space of drugs and chemical libraries

 setwd('c:/sci/037cinf/analysis/1')
 options(stringsAsFactors=FALSE)
 library(stringr)
 library(rcdk) # cheminformatics
 library(sp) # for point-in-polygon operations

 # http://www.cureffi.org/wp-content/uploads/2013/10/drugs.txt
 fda_drugs = read.table('c:/sci/037cinf/analysis/1/drugs.txt',sep='\t',header=TRUE)

 # created manually from published in vitro work
 # http://www.cureffi.org/wp-content/uploads/2013/11/antiprion_cpds.txt
 antiprion_cpds = read.table('c:/sci/037cinf/data/antiprion_cpds.txt',sep='\t',header=TRUE,comment.char='#',quote='')

 # load ZINC libraries screened 
 zinc_msspectrum_prop  = read.table('c:/sci/037cinf/data/msspectrum_prop.xls' ,header=TRUE,sep='\t',quote="",comment.char="")
 zinc_msusdrugs_prop   = read.table('c:/sci/037cinf/data/msusdrugs_prop.xls'  ,header=TRUE,sep='\t',quote="",comment.char="")
 zinc_smdccdivdiv_prop = read.table('c:/sci/037cinf/data/smdccdivdiv_prop.xls',header=TRUE,sep='\t',quote="",comment.char="")
 zinc_chbr_prop        = read.table('c:/sci/037cinf/data/chbr_prop.xls'       ,header=TRUE,sep='\t',quote="",comment.char="")

 # ChemBridge is too big, take a random subset of 20,000 rows
 # http://stackoverflow.com/questions/8273313/random-rows-in-dataframe-in-r
 set.seed(1234)
 zinc_chbr_prop_randsub = zinc_chbr_prop[sample(nrow(zinc_chbr_prop),20000,replace=FALSE),]
 rm(zinc_chbr_prop)
 # should check again - the matrix is not very wide, so maybe including all of ChemBridge _is_ feasible.

 # remove Inulin, a biological
 fda_drugs = subset(fda_drugs, generic_name != "Inulin")

 # set cpdset attribute to indicate which library or chemical set each molecule is part of
 fda_drugs$cpdset                     = "fda"
 fda_drugs$cpdset[fda_drugs$cns_drug] = "fda_cns"
 antiprion_cpds$cpdset = "antiprion"
 zinc_msspectrum_prop$cpdset   = "msspectrum"
 zinc_msusdrugs_prop$cpdset    = "msusdrugs"
 zinc_smdccdivdiv_prop$cpdset  = "smdccdivdiv"
 zinc_chbr_prop_randsub$cpdset = "chbr_randsub"

 # combine the zinc libraries
 zinclibs = rbind(zinc_msspectrum_prop,
            zinc_msusdrugs_prop,
            zinc_smdccdivdiv_prop,
            zinc_chbr_prop_randsub)

 # convert everything to same 3 columns - we just need compound set, compound name and SMILES for every molecule
 zinclibs = zinclibs[,c("cpdset","ZINC_ID","SMILES")]
 colnames(zinclibs) = c("cpdset","name","smiles")
 antiprion_cpds = antiprion_cpds[,c("cpdset","common_name","smiles")]
 colnames(antiprion_cpds) = c("cpdset","name","smiles")
 fda_drugs = fda_drugs[,c("cpdset","generic_name","smiles")]
 colnames(fda_drugs) = c("cpdset","name","smiles")

 # double check that names match
 colnames(zinclibs)
 colnames(antiprion_cpds)
 colnames(fda_drugs)

 # merge drugs, antiprion compounds and chemical libraries into one data frame
 mol = rbind(fda_drugs,antiprion_cpds,zinclibs)
 dim(mol)

 # get list of rcdk molecular descriptors, minus #28 IPLearning which makes R hang
 descriptors = get.desc.names(type='all')[-28]

 # create an empty results dataframe because rbinding new stuff in seems quicker
 # than indexing into a pre-sized df
 res = data.frame(id=integer(),name=character(),cpdset=character(),
                 smiles=character(),hba=integer(),hbd=integer(),nrb=integer(),
                 tpsa=numeric(),mwt=numeric(),xlogp=numeric(),charge=integer())

 # next block of code scarcely modified from "Properties of CNS drugs vs. all FDA-approved drugs"
 # note these indices are _after_ eliminating the original item #28
 descriptors_to_use = descriptors[c(25,26,39,41,45,48)] # respectively HBA, HBD, NRB, TPSA, MWT, XLOGP
 starttime = Sys.time()
 # test run: # for (i in 1:100) {
 for (i in 1:dim(mol)[1]) {
  if (mol$smiles[i] == "") { next }
  # try catch logic from http://stackoverflow.com/questions/8093914/skip-to-next-value-of-loop-upon-error-in-r-trycatch
  parse_smiles_error = tryCatch ( 
   { cpd_object = parse.smiles(mol$smiles[i]) },
     error = function(e) e
  )
  if(inherits(parse_smiles_error,"error")) next
  id = i
  hba = eval.desc(cpd_object, descriptors_to_use[1])
  hbd = eval.desc(cpd_object, descriptors_to_use[2])
  nrb = eval.desc(cpd_object, descriptors_to_use[3])
  tpsa = eval.desc(cpd_object, descriptors_to_use[4])
  mwt = eval.desc(cpd_object, descriptors_to_use[5])
  xlogp_error = tryCatch( 
    # throws null pointer exception on aminophylline
    { xlogp = eval.desc(cpd_object, descriptors_to_use[6]) },
      error = function(e) e
  )
  if(inherits(xlogp_error,"error")) next # just skip any molecules that throw null pointed exceptions on xlogp
  nplus = str_count(mol$smiles[i],"[+]") # + is not a metacharacter inside [] in POSIX regular expressions, see http://farazmahmood.wordpress.com/2008/11/10/regular-expression-escaping-meta-characters-within-character-classes/
  nminus = str_count(mol$smiles[i],"[-]")
  charge = nplus - nminus
  newrow = c(id,mol$name[i],mol$cpdset[i],mol$smiles[i],
             hba,hbd,nrb,tpsa,mwt,xlogp,charge)
  names(newrow) = colnames(res)
  res = rbind(res,newrow)
 }
 elapsed = Sys.time() - starttime
 elapsed

 # backup this effort so far in case principal component code makes R hang
 write.table(res,'res.txt',row.names=FALSE,col.names=FALSE,sep='\t',quote=FALSE)
 # re-read from disk
 res = read.table('res.txt',header=TRUE,sep='\t',comment.char='')
 colnames(res) = c("id","name","cpdset","smiles","hba","hbd","nrb","tpsa","mwt","xlogp","charge")


 ## NOTES ON PERFORMANCE
 # interesting - if you write the results into the original table with
 # indexing, e.g. mol$hba[i] = eval.desc(cpd_object, descriptors_to_use[1]),
 # this takes 24 seconds to do 10 rows and 3 mins to do 20 rows.
 # while if you write out to a new df and rbind as you go, it takes 1.1 seconds
 # to do 20 rows, and 4.9 seconds to do 100 rows
 # at 5 sec / 100 rows, should take 500 sec = ~ 8 minutes to do all.
 # instead, it got to i = 23306 in Time difference of 21.09367 mins
 # perhaps the chemical libraries are harder to do? or it started running out of memory?
 # if constant rate, whole thing will take over an hour

 # principal components

 mol_mat = as.matrix(res[,c('hba','hbd','nrb','tpsa','mwt','xlogp')]) #,'charge')])
 covmatrix = cov(mol_mat) # calculate covariance matrix
 eig = eigen(covmatrix) 
 pc = mol_mat %*% eig$vectors 
 var_explained = paste(formatC((eig$values/sum(eig$values))[1:3]*100,digits=1,format='f'),'%',sep='')
 var_explained
 colnames(pc) = paste("C",1:7,sep="")

 cpddata = cbind(res, pc)
 write.table(cpddata,'cpddata.txt',row.names=FALSE,col.names=TRUE,quote=FALSE,sep='\t')
 write.table(var_explained,'var_explained.txt',row.names=FALSE,col.names=TRUE,quote=FALSE,sep='\t')

 # break point: code above and below can be run independently.
 # code above takes ~ 2.5 hour, and below takes just a few mins


 cpddata = read.table('cpddata.txt',header=TRUE,sep='\t',quote='',comment.char='')
 var_explained = read.table('var_explained.txt',header=TRUE,sep='\t',quote='',comment.char='')
 var_explained = var_explained$x

 # list of annotations & plotting properties for the different libraries
 all_annot = data.frame(cpdset = c("fda","fda_cns","antiprion","msspectrum","msusdrugs","smdccdivdiv","chbr_randsub"),
                       desc   = c("All FDA-approved drugs","CNS drugs","Published anti-prion compounds","MSDI Spectrum Collection","MSDI US Drug Collection","UCSF SMDC ChemDiv Diversity","ChemBridge Random 20K subset"),
                       k      = c("#B3BDFF","#FCD477","#5E0000","#000000","#000000","#000000","#000000"),
                       pch    = c(19,19,17,15,16,17,18))

 # check names and rename principal component columns
 names(cpddata)


 cpddata = merge(cpddata,all_annot,by='cpdset')
 par(font=1) # turn off bold
 # plot the FDA and CNS drugs alone
 png('fda.cns.pc1.pc2.png',width=600,height=400)
 # grab annotations 
 lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns'),c('cpdset','desc','k','pch')])
 # first plot empty
 plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
     main = "FDA-approved drugs PC1 vs. PC2",
     cex.main = 1.2)
 # then plot different sets individually in order to control which is on top of others
 # FDA-approved drugs as base
 points(cpddata[cpddata$cpdset=='fda',"C1"],cpddata[cpddata$cpdset=='fda',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
 # CNS drugs as overlay
 points(cpddata[cpddata$cpdset=='fda_cns',"C1"],cpddata[cpddata$cpdset=='fda_cns',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
 # legend
 legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=1)

 # plot outliers
 outliers = which(  cpddata$C1 %in% c(min(cpddata$C1),max(cpddata$C1)) | 
                     cpddata$C2 %in% c(min(cpddata$C2)) ) # ,max(cpddata$C2) close to min(cpddata$C1)
 for (outlier in outliers) {
  points(cpddata[outlier,"C1"],cpddata[outlier,"C2"],pch=19,col='#000000')
  # plot text on appropriate side so it's visible
  if(cpddata[outlier,"C1"]==max(cpddata$C1)) {
    pos = 2
  } else {
    pos = 4
  }
  text(cpddata[outlier,"C1"],cpddata[outlier,"C2"],labels=cpddata[outlier,'name'],cex=.8,pos=pos)  
 }

 dev.off()

 # check on the leftmost cns drug
 which(cpddata[cpddata$cpdset=='fda_cns',"C1"]==min(cpddata[cpddata$cpdset=='fda_cns',"C1"]))
 res[180,] # bromocriptine, used in treating Parkinson's, so def CNS

 # here is a useful function
 # rcdk cannot plot disconnected molecules (e.g. w/ ionic bonds)
 # so this function will strip away the smaller disconnected parts
 stripsmiles = function(smiles) {
  parts = strsplit(smiles,"\\.")[[1]]
  longestpart = parts[which(nchar(parts)==max(nchar(parts)))]
  return(longestpart)
 }

 # size of range. if there's a base function for this I couldn't find it
 rangesize = function(vec) {
  return (max(vec) - min(vec))
 }

 # draw the outliers
 png('fda.outliers.png',width=600,height=250)
 par(mfrow=c(1,3),oma=c(0,0,3,0),mar=c(2,2,2,2))
 for (outlier in outliers) {
 molecule = parse.smiles(stripsmiles(cpddata$smiles[outlier]))[[1]]
 plot(NA,NA,xlim=c(1,10),ylim=c(1,10),xaxt='n',yaxt='n',
     xlab='',ylab='',main=cpddata$name[outlier])
 temp1 = view.image.2d(molecule,200,200) # convert Java representation to raster
 rasterImage(temp1,1,1,10,10)
 }
 dev.off()


 # list of annotations & plotting properties for the different libraries
 all_annot = data.frame(cpdset = c("fda","fda_cns","antiprion","msspectrum","msusdrugs","smdccdivdiv","chbr_randsub"),
                       desc   = c("All FDA-approved drugs","CNS drugs","Published anti-prion compounds","MSDI Spectrum Collection","MSDI US Drug Collection","UCSF SMDC ChemDiv Diversity","ChemBridge Random 20K subset"),
                       k      = c("#B3BDFF","#FCD477","#5E0000","#000000","#000000","#000000","#000000"),
                       pch    = c(19,19,17,15,16,17,18))



 # 4-panel plot of each compound library's overlap with the drug / CNS drug space
 png('chemlibs.pc1.pc2.overlay.png',width=900,height=600)
 par(mfrow=c(2,2),oma=c(0,0,3,0),mar=c(4,4,4,4))
 for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
  # grab annotations for current library
  lib_annot = unique(cpddata[cpddata$cpdset %in% c('fda','fda_cns',chemlib),c('cpdset','desc','k','pch')])
  
  # first plot empty
  plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
     main = lib_annot$desc[lib_annot$cpdset==chemlib],
     cex.main = 1.2)
  # then plot different sets individually in order to control which is on top of others
  # FDA-approved drugs as base
  points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
  # CNS drugs as overlay
  points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
  # then all chemical libraries as smaller points on top
  points(cpddata$C1[cpddata$cpdset==chemlib],cpddata$C2[cpddata$cpdset==chemlib],pch=cpddata$pch[cpddata$cpdset==chemlib],cex=.6,col=cpddata$k[cpddata$cpdset==chemlib])  #col=res$k[res$cpdset==chemlib])
  
  # don't overlay the convex hulls
 #  points(drug_pts[drug_chull_idx,"C1"],drug_pts[drug_chull_idx,"C2"],type='l',lwd=3,col='blue')
 #  points(cns_pts[cns_chull_idx,"C1"],cns_pts[cns_chull_idx,"C2"],type='l',lwd=3,col='orange')
  
  # legend
  legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)
 }
 mtext("Drugs and chemical libraries PC1 vs. PC2", outer = TRUE, cex = 1.5)
 dev.off()

 png('chemlibs.pc1.pc2.mwt.hist.png',width=900,height=600)
 par(mfrow=c(2,2),oma=c(0,0,3,0),mar=c(4,4,4,4))
 for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
  hist(cpddata$mwt[cpddata$cpdset==chemlib],xlim=c(0,1000),col='black',
       xlab='Molecular weight',
       main=all_annot$desc[all_annot$cpdset==chemlib])
 }
 mtext("Molecular weight distribution of chemical libraries", outer = TRUE, cex = 1.5)
 dev.off()



 # plot the FDA and CNS drugs with their convex hulls
 png('fda.cns.pc1.pc2.chulls.png',width=600,height=400)
 # grab annotations 
 lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns'),c('cpdset','desc','k','pch')])
 # first plot empty
 plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
     main = "FDA-approved drugs PC1 vs. PC2",
     cex.main = 1.2)
 # scatterplot the FDA-approved drugs
 points(cpddata[cpddata$cpdset=='fda',"C1"],cpddata[cpddata$cpdset=='fda',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
 # scatterplot the CNS drugs
 points(cpddata[cpddata$cpdset=='fda_cns',"C1"],cpddata[cpddata$cpdset=='fda_cns',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])

 # find convex hull for all drugs
 drug_pts = cpddata[cpddata$cpdset %in% c('fda','fda_cns'),]
 drug_chull_idx = chull(drug_pts$C1,drug_pts$C2)
 points(drug_pts[drug_chull_idx,"C1"],drug_pts[drug_chull_idx,"C2"],type='l',lwd=3,col='blue')

 # find the convex hull of the CNS drugs
 cns_pts = cpddata[cpddata$cpdset=='fda_cns',]
 cns_chull_idx = chull(cns_pts$C1,cns_pts$C2)
 points(cns_pts[cns_chull_idx,"C1"],cns_pts[cns_chull_idx,"C2"],type='l',lwd=3,col='orange')

 # legend
 legend('topright',c("Drug space","CNS drug space"),col=c('blue','orange'),lwd=3)

 dev.off()



 # table to hold convex hull results
 chtable = data.frame(cpdset=character(),desc=character(),cns_frac=numeric(),drug_frac=numeric())
 # now compute fraction within the CNS polygon for all libraries
 for (chemlib in all_annot$cpdset) {
  in_cns = point.in.polygon(cpddata[cpddata$cpdset==chemlib,"C1"],
                            cpddata[cpddata$cpdset==chemlib,"C2"],
                            cns_pts[cns_chull_idx,"C1"],
                            cns_pts[cns_chull_idx,"C2"])  
  in_drugs = point.in.polygon(cpddata[cpddata$cpdset==chemlib,"C1"],
                              cpddata[cpddata$cpdset==chemlib,"C2"],
                              drug_pts[drug_chull_idx,"C1"],
                              drug_pts[drug_chull_idx,"C2"])  
  temp1 = cbind(chemlib,
                all_annot$desc[all_annot$cpdset==chemlib],
                sum(in_cns > 0)/length(in_cns),
                sum(in_drugs > 0)/length(in_drugs))
  names(temp1) = c('cpdset','desc','cns_frac','drug_frac')
  chtable = rbind(chtable,temp1)
 }
 chtable


 dev.off()
 png('druglike.barplot.png',width=600,height=500)
 par(mar=c(10,5,5,5),font=2)
 chtable$cns_frac = as.numeric(chtable$V3)
 chtable$drug_addl = as.numeric(chtable$V4)-as.numeric(chtable$V3)
 plot.idx = c(1,4,5,6,7)
 bp = barplot(rbind(chtable$cns_frac[plot.idx],chtable$drug_addl[plot.idx]),beside=FALSE,
        names.arg=chtable$desc[plot.idx],col=c("#FCD477","#B3BDFF"),border=NA,
        yaxt='n',xaxt='n',
        main="Drug-like content of chemical libraries",
        ylab="Fraction of drugs in library in given chemical space")
 axis(side=2,at=seq(0,1,.2),labels=paste(seq(0,100,20),"%",sep=""))
 text(bp, par("usr")[3], labels = all_annot$desc[plot.idx], srt = 45, adj = c(1.1,1.1), xpd = TRUE, cex=.9)
 abline(h=0)
 abline(h=1)
 legend('bottomright',c("In CNS drug space","In all drug space"),pch=16,col=c("#FCD477","#B3BDFF"))
 dev.off()


 # now plot overlay with antiprion cpds
 png('drugs.vs.antiprion.png',width=600,height=400)
 # grab annotations  
 lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns','antiprion'),c('cpdset','desc','k','pch')])

 # first plot empty
 plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
     main = "Antiprion compounds in chemical space",
     cex.main = 1.2)
 # then plot different sets individually in order to control which is on top of others
 # FDA-approved drugs as base
 points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
 # CNS drugs as overlay
 points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
 # then all chemical libraries as smaller points on top
 #for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
 #  points(pc[res$cpdset==chemlib,1],pc[res$cpdset==chemlib,2],pch=res$pch[res$cpdset==chemlib],cex=.6,col='black')  #col=res$k[res$cpdset==chemlib])
 #}
 # finally, antiprion cpds 
 points(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],pch=cpddata$pch[cpddata$cpdset=='antiprion'],cex=1.5,col=cpddata$k[cpddata$cpdset=='antiprion'])

 par(font=2)
 text(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],labels=cpddata$name[cpddata$cpdset=='antiprion'],pos=4,col=cpddata$k[cpddata$cpdset=='antiprion'],cex=.8)
 # legend
 legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)

 dev.off()



 # re-plot zoomed in on the CNS drugs
 # now plot overlay with antiprion cpds
 png('drugs.vs.antiprion.zoomin.png',width=600,height=400)

 lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns','antiprion'),c('cpdset','desc','k','pch')])
 plot(NA,NA,xlim=range(cpddata$C1[cpddata$cpdset=='fda_cns']),ylim=range(cpddata$C2[cpddata$cpdset=='fda_cns']),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
     main = "Antiprion compounds in chemical space - CNS cluster",
     cex.main = 1.2)
 points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
 points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
 points(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],pch=cpddata$pch[cpddata$cpdset=='antiprion'],cex=1.5,col=cpddata$k[cpddata$cpdset=='antiprion'])


 # put labels on opposite sides in dense clusters
 cpd_subset = cpddata[cpddata$cpdset=='antiprion',]
 pos = rep(4,dim(cpd_subset)[1]) # default is right
 pos[cpd_subset$name=='tilorone'] = 2
 pos[cpd_subset$name=='arylamide 17'] = 2
 pos[cpd_subset$name=='arylpiperazine 41'] = 2
 par(font=2)
 text(cpd_subset[cpd_subset$cpdset=='antiprion',"C1"],cpd_subset[cpd_subset$cpdset=='antiprion',"C2"],labels=cpd_subset$name[cpd_subset$cpdset=='antiprion'],pos=pos,col=cpd_subset$k[cpd_subset$cpdset=='antiprion'],cex=.8)
 # legend
 legend('bottomright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)

 dev.off()




 # to browse the available pch offline:
 # par(mfrow=c(5,5),mar=c(0,0,0,0))
 # for(i in 1:25) { plot(1,1,pch=i,cex=3,xaxt='n',yaxt='n'); text(1,1.2,labels=i) }



 # plots:
 # 1. drugs with labels for outliers
 # 2. comparison of libraries with CNS and drugs
 #    discussion of this vs. literature result
 # 3. antiprion cpds, room for optimization


 # condense to a small table for legend
	# Eric Minikel
	# CureFFI.org
	# 2013-11-06
	# Anti-prion compounds plotted in the space of drugs and chemical libraries

	setwd('c:/sci/037cinf/analysis/1')
	options(stringsAsFactors=FALSE)
	library(stringr)
	library(rcdk) # cheminformatics
	library(sp) # for point-in-polygon operations

	# http://www.cureffi.org/wp-content/uploads/2013/10/drugs.txt
	fda_drugs = read.table('c:/sci/037cinf/analysis/1/drugs.txt',sep='\t',header=TRUE)

	# created manually from published in vitro work
	# http://www.cureffi.org/wp-content/uploads/2013/11/antiprion_cpds.txt
	antiprion_cpds = read.table('c:/sci/037cinf/data/antiprion_cpds.txt',sep='\t',header=TRUE,comment.char='#',quote='')

	# load ZINC libraries screened
	zinc_msspectrum_prop = read.table('c:/sci/037cinf/data/msspectrum_prop.xls' ,header=TRUE,sep='\t',quote="",comment.char="")
	zinc_msusdrugs_prop = read.table('c:/sci/037cinf/data/msusdrugs_prop.xls' ,header=TRUE,sep='\t',quote="",comment.char="")
	zinc_smdccdivdiv_prop = read.table('c:/sci/037cinf/data/smdccdivdiv_prop.xls',header=TRUE,sep='\t',quote="",comment.char="")
	zinc_chbr_prop = read.table('c:/sci/037cinf/data/chbr_prop.xls' ,header=TRUE,sep='\t',quote="",comment.char="")

	# ChemBridge is too big, take a random subset of 20,000 rows
	# http://stackoverflow.com/questions/8273313/random-rows-in-dataframe-in-r
	set.seed(1234)
	zinc_chbr_prop_randsub = zinc_chbr_prop[sample(nrow(zinc_chbr_prop),20000,replace=FALSE),]
	rm(zinc_chbr_prop)
	# should check again - the matrix is not very wide, so maybe including all of ChemBridge _is_ feasible.

	# remove Inulin, a biological
	fda_drugs = subset(fda_drugs, generic_name != "Inulin")

	# set cpdset attribute to indicate which library or chemical set each molecule is part of
	fda_drugs$cpdset = "fda"
	fda_drugs$cpdset[fda_drugs$cns_drug] = "fda_cns"
	antiprion_cpds$cpdset = "antiprion"
	zinc_msspectrum_prop$cpdset = "msspectrum"
	zinc_msusdrugs_prop$cpdset = "msusdrugs"
	zinc_smdccdivdiv_prop$cpdset = "smdccdivdiv"
	zinc_chbr_prop_randsub$cpdset = "chbr_randsub"

	# combine the zinc libraries
	zinclibs = rbind(zinc_msspectrum_prop,
	zinc_msusdrugs_prop,
	zinc_smdccdivdiv_prop,
	zinc_chbr_prop_randsub)

	# convert everything to same 3 columns - we just need compound set, compound name and SMILES for every molecule
	zinclibs = zinclibs[,c("cpdset","ZINC_ID","SMILES")]
	colnames(zinclibs) = c("cpdset","name","smiles")
	antiprion_cpds = antiprion_cpds[,c("cpdset","common_name","smiles")]
	colnames(antiprion_cpds) = c("cpdset","name","smiles")
	fda_drugs = fda_drugs[,c("cpdset","generic_name","smiles")]
	colnames(fda_drugs) = c("cpdset","name","smiles")

	# double check that names match
	colnames(zinclibs)
	colnames(antiprion_cpds)
	colnames(fda_drugs)

	# merge drugs, antiprion compounds and chemical libraries into one data frame
	mol = rbind(fda_drugs,antiprion_cpds,zinclibs)
	dim(mol)

	# get list of rcdk molecular descriptors, minus #28 IPLearning which makes R hang
	descriptors = get.desc.names(type='all')[-28]

	# create an empty results dataframe because rbinding new stuff in seems quicker
	# than indexing into a pre-sized df
	res = data.frame(id=integer(),name=character(),cpdset=character(),
	smiles=character(),hba=integer(),hbd=integer(),nrb=integer(),
	tpsa=numeric(),mwt=numeric(),xlogp=numeric(),charge=integer())

	# next block of code scarcely modified from "Properties of CNS drugs vs. all FDA-approved drugs"
	# note these indices are _after_ eliminating the original item #28
	descriptors_to_use = descriptors[c(25,26,39,41,45,48)] # respectively HBA, HBD, NRB, TPSA, MWT, XLOGP
	starttime = Sys.time()
	# test run: # for (i in 1:100) {
	for (i in 1:dim(mol)[1]) {
	if (mol$smiles[i] == "") { next }
	# try catch logic from http://stackoverflow.com/questions/8093914/skip-to-next-value-of-loop-upon-error-in-r-trycatch
	parse_smiles_error = tryCatch (
	{ cpd_object = parse.smiles(mol$smiles[i]) },
	error = function(e) e
	)
	if(inherits(parse_smiles_error,"error")) next
	id = i
	hba = eval.desc(cpd_object, descriptors_to_use[1])
	hbd = eval.desc(cpd_object, descriptors_to_use[2])
	nrb = eval.desc(cpd_object, descriptors_to_use[3])
	tpsa = eval.desc(cpd_object, descriptors_to_use[4])
	mwt = eval.desc(cpd_object, descriptors_to_use[5])
	xlogp_error = tryCatch(
	# throws null pointer exception on aminophylline
	{ xlogp = eval.desc(cpd_object, descriptors_to_use[6]) },
	error = function(e) e
	)
	if(inherits(xlogp_error,"error")) next # just skip any molecules that throw null pointed exceptions on xlogp
	nplus = str_count(mol$smiles[i],"[+]") # + is not a metacharacter inside [] in POSIX regular expressions, see http://farazmahmood.wordpress.com/2008/11/10/regular-expression-escaping-meta-characters-within-character-classes/
	nminus = str_count(mol$smiles[i],"[-]")
	charge = nplus - nminus
	newrow = c(id,mol$name[i],mol$cpdset[i],mol$smiles[i],
	hba,hbd,nrb,tpsa,mwt,xlogp,charge)
	names(newrow) = colnames(res)
	res = rbind(res,newrow)
	}
	elapsed = Sys.time() - starttime
	elapsed

	# backup this effort so far in case principal component code makes R hang
	write.table(res,'res.txt',row.names=FALSE,col.names=FALSE,sep='\t',quote=FALSE)
	# re-read from disk
	res = read.table('res.txt',header=TRUE,sep='\t',comment.char='')
	colnames(res) = c("id","name","cpdset","smiles","hba","hbd","nrb","tpsa","mwt","xlogp","charge")


	## NOTES ON PERFORMANCE
	# interesting - if you write the results into the original table with
	# indexing, e.g. mol$hba[i] = eval.desc(cpd_object, descriptors_to_use[1]),
	# this takes 24 seconds to do 10 rows and 3 mins to do 20 rows.
	# while if you write out to a new df and rbind as you go, it takes 1.1 seconds
	# to do 20 rows, and 4.9 seconds to do 100 rows
	# at 5 sec / 100 rows, should take 500 sec = ~ 8 minutes to do all.
	# instead, it got to i = 23306 in Time difference of 21.09367 mins
	# perhaps the chemical libraries are harder to do? or it started running out of memory?
	# if constant rate, whole thing will take over an hour

	# principal components

	mol_mat = as.matrix(res[,c('hba','hbd','nrb','tpsa','mwt','xlogp')]) #,'charge')])
	covmatrix = cov(mol_mat) # calculate covariance matrix
	eig = eigen(covmatrix)
	pc = mol_mat %*% eig$vectors
	var_explained = paste(formatC((eig$values/sum(eig$values))[1:3]*100,digits=1,format='f'),'%',sep='')
	var_explained
	colnames(pc) = paste("C",1:7,sep="")

	cpddata = cbind(res, pc)
	write.table(cpddata,'cpddata.txt',row.names=FALSE,col.names=TRUE,quote=FALSE,sep='\t')
	write.table(var_explained,'var_explained.txt',row.names=FALSE,col.names=TRUE,quote=FALSE,sep='\t')

	# break point: code above and below can be run independently.
	# code above takes ~ 2.5 hour, and below takes just a few mins


	cpddata = read.table('cpddata.txt',header=TRUE,sep='\t',quote='',comment.char='')
	var_explained = read.table('var_explained.txt',header=TRUE,sep='\t',quote='',comment.char='')
	var_explained = var_explained$x

	# list of annotations & plotting properties for the different libraries
	all_annot = data.frame(cpdset = c("fda","fda_cns","antiprion","msspectrum","msusdrugs","smdccdivdiv","chbr_randsub"),
	desc = c("All FDA-approved drugs","CNS drugs","Published anti-prion compounds","MSDI Spectrum Collection","MSDI US Drug Collection","UCSF SMDC ChemDiv Diversity","ChemBridge Random 20K subset"),
	k = c("#B3BDFF","#FCD477","#5E0000","#000000","#000000","#000000","#000000"),
	pch = c(19,19,17,15,16,17,18))

	# check names and rename principal component columns
	names(cpddata)


	cpddata = merge(cpddata,all_annot,by='cpdset')
	par(font=1) # turn off bold
	# plot the FDA and CNS drugs alone
	png('fda.cns.pc1.pc2.png',width=600,height=400)
	# grab annotations
	lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns'),c('cpdset','desc','k','pch')])
	# first plot empty
	plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
	main = "FDA-approved drugs PC1 vs. PC2",
	cex.main = 1.2)
	# then plot different sets individually in order to control which is on top of others
	# FDA-approved drugs as base
	points(cpddata[cpddata$cpdset=='fda',"C1"],cpddata[cpddata$cpdset=='fda',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
	# CNS drugs as overlay
	points(cpddata[cpddata$cpdset=='fda_cns',"C1"],cpddata[cpddata$cpdset=='fda_cns',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
	# legend
	legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=1)

	# plot outliers
	outliers = which( cpddata$C1 %in% c(min(cpddata$C1),max(cpddata$C1)) \|
	cpddata$C2 %in% c(min(cpddata$C2)) ) # ,max(cpddata$C2) close to min(cpddata$C1)
	for (outlier in outliers) {
	points(cpddata[outlier,"C1"],cpddata[outlier,"C2"],pch=19,col='#000000')
	# plot text on appropriate side so it's visible
	if(cpddata[outlier,"C1"]==max(cpddata$C1)) {
	pos = 2
	} else {
	pos = 4
	}
	text(cpddata[outlier,"C1"],cpddata[outlier,"C2"],labels=cpddata[outlier,'name'],cex=.8,pos=pos)
	}

	dev.off()

	# check on the leftmost cns drug
	which(cpddata[cpddata$cpdset=='fda_cns',"C1"]==min(cpddata[cpddata$cpdset=='fda_cns',"C1"]))
	res[180,] # bromocriptine, used in treating Parkinson's, so def CNS

	# here is a useful function
	# rcdk cannot plot disconnected molecules (e.g. w/ ionic bonds)
	# so this function will strip away the smaller disconnected parts
	stripsmiles = function(smiles) {
	parts = strsplit(smiles,"\\.")[[1]]
	longestpart = parts[which(nchar(parts)==max(nchar(parts)))]
	return(longestpart)
	}

	# size of range. if there's a base function for this I couldn't find it
	rangesize = function(vec) {
	return (max(vec) - min(vec))
	}

	# draw the outliers
	png('fda.outliers.png',width=600,height=250)
	par(mfrow=c(1,3),oma=c(0,0,3,0),mar=c(2,2,2,2))
	for (outlier in outliers) {
	molecule = parse.smiles(stripsmiles(cpddata$smiles[outlier]))[[1]]
	plot(NA,NA,xlim=c(1,10),ylim=c(1,10),xaxt='n',yaxt='n',
	xlab='',ylab='',main=cpddata$name[outlier])
	temp1 = view.image.2d(molecule,200,200) # convert Java representation to raster
	rasterImage(temp1,1,1,10,10)
	}
	dev.off()


	# list of annotations & plotting properties for the different libraries
	all_annot = data.frame(cpdset = c("fda","fda_cns","antiprion","msspectrum","msusdrugs","smdccdivdiv","chbr_randsub"),
	desc = c("All FDA-approved drugs","CNS drugs","Published anti-prion compounds","MSDI Spectrum Collection","MSDI US Drug Collection","UCSF SMDC ChemDiv Diversity","ChemBridge Random 20K subset"),
	k = c("#B3BDFF","#FCD477","#5E0000","#000000","#000000","#000000","#000000"),
	pch = c(19,19,17,15,16,17,18))



	# 4-panel plot of each compound library's overlap with the drug / CNS drug space
	png('chemlibs.pc1.pc2.overlay.png',width=900,height=600)
	par(mfrow=c(2,2),oma=c(0,0,3,0),mar=c(4,4,4,4))
	for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
	# grab annotations for current library
	lib_annot = unique(cpddata[cpddata$cpdset %in% c('fda','fda_cns',chemlib),c('cpdset','desc','k','pch')])

	# first plot empty
	plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
	main = lib_annot$desc[lib_annot$cpdset==chemlib],
	cex.main = 1.2)
	# then plot different sets individually in order to control which is on top of others
	# FDA-approved drugs as base
	points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
	# CNS drugs as overlay
	points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
	# then all chemical libraries as smaller points on top
	points(cpddata$C1[cpddata$cpdset==chemlib],cpddata$C2[cpddata$cpdset==chemlib],pch=cpddata$pch[cpddata$cpdset==chemlib],cex=.6,col=cpddata$k[cpddata$cpdset==chemlib]) #col=res$k[res$cpdset==chemlib])

	# don't overlay the convex hulls
	# points(drug_pts[drug_chull_idx,"C1"],drug_pts[drug_chull_idx,"C2"],type='l',lwd=3,col='blue')
	# points(cns_pts[cns_chull_idx,"C1"],cns_pts[cns_chull_idx,"C2"],type='l',lwd=3,col='orange')

	# legend
	legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)
	}
	mtext("Drugs and chemical libraries PC1 vs. PC2", outer = TRUE, cex = 1.5)
	dev.off()

	png('chemlibs.pc1.pc2.mwt.hist.png',width=900,height=600)
	par(mfrow=c(2,2),oma=c(0,0,3,0),mar=c(4,4,4,4))
	for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
	hist(cpddata$mwt[cpddata$cpdset==chemlib],xlim=c(0,1000),col='black',
	xlab='Molecular weight',
	main=all_annot$desc[all_annot$cpdset==chemlib])
	}
	mtext("Molecular weight distribution of chemical libraries", outer = TRUE, cex = 1.5)
	dev.off()



	# plot the FDA and CNS drugs with their convex hulls
	png('fda.cns.pc1.pc2.chulls.png',width=600,height=400)
	# grab annotations
	lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns'),c('cpdset','desc','k','pch')])
	# first plot empty
	plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
	main = "FDA-approved drugs PC1 vs. PC2",
	cex.main = 1.2)
	# scatterplot the FDA-approved drugs
	points(cpddata[cpddata$cpdset=='fda',"C1"],cpddata[cpddata$cpdset=='fda',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
	# scatterplot the CNS drugs
	points(cpddata[cpddata$cpdset=='fda_cns',"C1"],cpddata[cpddata$cpdset=='fda_cns',"C2"],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])

	# find convex hull for all drugs
	drug_pts = cpddata[cpddata$cpdset %in% c('fda','fda_cns'),]
	drug_chull_idx = chull(drug_pts$C1,drug_pts$C2)
	points(drug_pts[drug_chull_idx,"C1"],drug_pts[drug_chull_idx,"C2"],type='l',lwd=3,col='blue')

	# find the convex hull of the CNS drugs
	cns_pts = cpddata[cpddata$cpdset=='fda_cns',]
	cns_chull_idx = chull(cns_pts$C1,cns_pts$C2)
	points(cns_pts[cns_chull_idx,"C1"],cns_pts[cns_chull_idx,"C2"],type='l',lwd=3,col='orange')

	# legend
	legend('topright',c("Drug space","CNS drug space"),col=c('blue','orange'),lwd=3)

	dev.off()



	# table to hold convex hull results
	chtable = data.frame(cpdset=character(),desc=character(),cns_frac=numeric(),drug_frac=numeric())
	# now compute fraction within the CNS polygon for all libraries
	for (chemlib in all_annot$cpdset) {
	in_cns = point.in.polygon(cpddata[cpddata$cpdset==chemlib,"C1"],
	cpddata[cpddata$cpdset==chemlib,"C2"],
	cns_pts[cns_chull_idx,"C1"],
	cns_pts[cns_chull_idx,"C2"])
	in_drugs = point.in.polygon(cpddata[cpddata$cpdset==chemlib,"C1"],
	cpddata[cpddata$cpdset==chemlib,"C2"],
	drug_pts[drug_chull_idx,"C1"],
	drug_pts[drug_chull_idx,"C2"])
	temp1 = cbind(chemlib,
	all_annot$desc[all_annot$cpdset==chemlib],
	sum(in_cns > 0)/length(in_cns),
	sum(in_drugs > 0)/length(in_drugs))
	names(temp1) = c('cpdset','desc','cns_frac','drug_frac')
	chtable = rbind(chtable,temp1)
	}
	chtable


	dev.off()
	png('druglike.barplot.png',width=600,height=500)
	par(mar=c(10,5,5,5),font=2)
	chtable$cns_frac = as.numeric(chtable$V3)
	chtable$drug_addl = as.numeric(chtable$V4)-as.numeric(chtable$V3)
	plot.idx = c(1,4,5,6,7)
	bp = barplot(rbind(chtable$cns_frac[plot.idx],chtable$drug_addl[plot.idx]),beside=FALSE,
	names.arg=chtable$desc[plot.idx],col=c("#FCD477","#B3BDFF"),border=NA,
	yaxt='n',xaxt='n',
	main="Drug-like content of chemical libraries",
	ylab="Fraction of drugs in library in given chemical space")
	axis(side=2,at=seq(0,1,.2),labels=paste(seq(0,100,20),"%",sep=""))
	text(bp, par("usr")[3], labels = all_annot$desc[plot.idx], srt = 45, adj = c(1.1,1.1), xpd = TRUE, cex=.9)
	abline(h=0)
	abline(h=1)
	legend('bottomright',c("In CNS drug space","In all drug space"),pch=16,col=c("#FCD477","#B3BDFF"))
	dev.off()


	# now plot overlay with antiprion cpds
	png('drugs.vs.antiprion.png',width=600,height=400)
	# grab annotations
	lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns','antiprion'),c('cpdset','desc','k','pch')])

	# first plot empty
	plot(NA,NA,xlim=range(cpddata$C1),ylim=range(cpddata$C2),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
	main = "Antiprion compounds in chemical space",
	cex.main = 1.2)
	# then plot different sets individually in order to control which is on top of others
	# FDA-approved drugs as base
	points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
	# CNS drugs as overlay
	points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
	# then all chemical libraries as smaller points on top
	#for (chemlib in c("msspectrum","msusdrugs","smdccdivdiv","chbr_randsub")) {
	# points(pc[res$cpdset==chemlib,1],pc[res$cpdset==chemlib,2],pch=res$pch[res$cpdset==chemlib],cex=.6,col='black') #col=res$k[res$cpdset==chemlib])
	#}
	# finally, antiprion cpds
	points(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],pch=cpddata$pch[cpddata$cpdset=='antiprion'],cex=1.5,col=cpddata$k[cpddata$cpdset=='antiprion'])

	par(font=2)
	text(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],labels=cpddata$name[cpddata$cpdset=='antiprion'],pos=4,col=cpddata$k[cpddata$cpdset=='antiprion'],cex=.8)
	# legend
	legend('topright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)

	dev.off()



	# re-plot zoomed in on the CNS drugs
	# now plot overlay with antiprion cpds
	png('drugs.vs.antiprion.zoomin.png',width=600,height=400)

	lib_annot = unique(all_annot[all_annot$cpdset %in% c('fda','fda_cns','antiprion'),c('cpdset','desc','k','pch')])
	plot(NA,NA,xlim=range(cpddata$C1[cpddata$cpdset=='fda_cns']),ylim=range(cpddata$C2[cpddata$cpdset=='fda_cns']),xlab=paste('PC1 (',var_explained[1],')',sep=''),ylab=paste('PC2 (',var_explained[2],')',sep=''),
	main = "Antiprion compounds in chemical space - CNS cluster",
	cex.main = 1.2)
	points(cpddata$C1[cpddata$cpdset=='fda'],cpddata$C2[cpddata$cpdset=='fda'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda'])
	points(cpddata$C1[cpddata$cpdset=='fda_cns'],cpddata$C2[cpddata$cpdset=='fda_cns'],pch=19,cex=3,col=cpddata$k[cpddata$cpdset=='fda_cns'])
	points(cpddata$C1[cpddata$cpdset=='antiprion'],cpddata$C2[cpddata$cpdset=='antiprion'],pch=cpddata$pch[cpddata$cpdset=='antiprion'],cex=1.5,col=cpddata$k[cpddata$cpdset=='antiprion'])


	# put labels on opposite sides in dense clusters
	cpd_subset = cpddata[cpddata$cpdset=='antiprion',]
	pos = rep(4,dim(cpd_subset)[1]) # default is right
	pos[cpd_subset$name=='tilorone'] = 2
	pos[cpd_subset$name=='arylamide 17'] = 2
	pos[cpd_subset$name=='arylpiperazine 41'] = 2
	par(font=2)
	text(cpd_subset[cpd_subset$cpdset=='antiprion',"C1"],cpd_subset[cpd_subset$cpdset=='antiprion',"C2"],labels=cpd_subset$name[cpd_subset$cpdset=='antiprion'],pos=pos,col=cpd_subset$k[cpd_subset$cpdset=='antiprion'],cex=.8)
	# legend
	legend('bottomright',lib_annot$desc,col=lib_annot$k,pch=lib_annot$pch,cex=.6)

	dev.off()




	# to browse the available pch offline:
	# par(mfrow=c(5,5),mar=c(0,0,0,0))
	# for(i in 1:25) { plot(1,1,pch=i,cex=3,xaxt='n',yaxt='n'); text(1,1.2,labels=i) }



	# plots:
	# 1. drugs with labels for outliers
	# 2. comparison of libraries with CNS and drugs
	# discussion of this vs. literature result
	# 3. antiprion cpds, room for optimization


	# condense to a small table for legend