R/getChemSimNet.R

In kwanjeeraw/metabox: metabox

#compute similarity network, D. Barupal
getChemSimNet <- function (cids, cutoff=0.7) {
  #cids <- c(1:5)
  #cutoff <- 0.7
  library(RCurl)
  hex <- c("0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "a", "b", "c", "d", "e", "f")
  bin <- c("0000", "0001", "0010", "0011", "0100", "0101", "0110", "0111", "1000", "1001", "1010", "1011", "1100", "1101", "1110", "1111")
  subkeys <- read.csv(paste(c('https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/cid/',paste(cids,collapse=","),'/property/Fingerprint2D/csv'),collapse=""))
  m <- t(sapply(subkeys[,2],function(x) { as.integer(strsplit(paste(sapply(strsplit(paste(base64Decode(x,"raw")[5:115],collapse=""),"")[[1]],function(x) {bin[which(hex==x)]}),collapse=""),"")[[1]][1:881] ) }))
  mat <- m%*%t(m)
  len <- length(m[,1])
  s <- mat.or.vec(len,len)
  for (i in 1:len) {
    for (j in 1:len){
      s[i,j] <- mat[i,j]/(mat[i,i]+mat[j,j]-mat[i,j])
    }
  }
  diag(s) <- 0
  s[lower.tri(s)]<-0
  return(do.call(rbind,sapply(1:length(cids), function (k) { if(length(which(s[k,]>cutoff))>0) {cbind(cids[k],cids[which(s[k,]>cutoff)],s[k,][which(s[k,]>cutoff)])}} )))
}