R/repelPoints.R

In PoisonAlien/maftools: Summarize, Analyze and Visualize MAF Files

#Repel points (shift points 40% to their inter cluster distance.)
shiftPoints = function(clust, leftDist, rightDist){
  
  if(nrow(clust) == 1){
    clust$pos2 = clust$pos
    return(clust)
  } else if(nrow(clust) == 2){
    #if distance is huge between right and left clusters, bring it down to 50.
    leftDist = ifelse(test = leftDist >100, yes = 50, no = leftDist )
    rightDist = ifelse(test = rightDist >100, yes = 50, no = rightDist)
    
    startPoint = clust[1,pos] - as.integer(leftDist * 0.2)
    endPoint = clust[nrow(clust), pos] + as.integer(rightDist * 0.2)  
    clust$pos2 = c(startPoint, endPoint)
    return(clust)
  } else{
    
    leftDist = ifelse(test = leftDist >100, yes =  50, no = leftDist)
    rightDist = ifelse(test = rightDist >100, yes = 50, no = rightDist)
    
    startPoint = clust[1,pos] - as.integer(leftDist * 0.4)
    endPoint = clust[nrow(clust), pos] + as.integer(rightDist * 0.4)
    posPoints = clust[2:(nrow(clust)-1), pos]  
    
    dist = rbind(abs(posPoints - startPoint), abs(posPoints - endPoint))
    dist = sapply(apply(dist, 2, function(x) which(x == max(x))), '[[', 1)
    
    
    tempStart = startPoint
    tempEnd = endPoint
    
    for(i in 1:length(posPoints)){
      
      if(dist[i] == 1){
        posPoints[i] = tempStart + as.integer(0.4 * abs(tempStart - posPoints[i]))
        tempStart = posPoints[i]  
      }else{
        posPoints[i] = tempEnd - as.integer(0.4 * abs(tempEnd - posPoints[i]))
        tempEnd = posPoints[i]
      }
    }
    clust$pos2 = c(startPoint, posPoints, endPoint)
    return(clust)
  }
}

#cluster
repelPoints = function(dat, protLen, clustSize){
  
  dat = data.table(dat)
  
  dat = dat[order(dat$pos),]
  dat$distance = c(0,diff(dat$pos))
  
  ## cluster identification
  mergeDist = clustSize #hard coded inter event distance.
  cdf = c()
  cluster = 1
  n = 1
  start = end = dat[1,pos]
  
  for(i in 2:nrow(dat)){
    
    dist = dat[i,distance]
    position = dat[i,pos]
    
    if(dist <= mergeDist){
      end = position
      n = n+1
    }else{
      tempdf = data.frame(cluster = paste('cluster', cluster, sep='_'), start = start, end = end, n = n )
      n = 1
      cdf = rbind(cdf, tempdf)
      cluster = cluster + 1
      start = end = position
    }
  }
  cdf = rbind(cdf, data.frame(cluster = paste('cluster', cluster, sep='_'), start = start, end = end, n = n))
  
  #Assign clusters to main table
  cdf2 = c()
  for(i in 1:nrow(cdf)){
    start = cdf[i,"start"]
    end = cdf[i,"end"]
    tdat = dat[pos >= start & pos <= end]
    tdat[,cluster:= paste('cluster', i, sep='_')]
    cdf2 = rbind(cdf2, tdat)
  }
  
  #Cluster summary and inter cluster distance
  clustSum = cdf2[,.(min = min(pos), max = max(pos)), by = cluster]
  clustEnd = clustSum[1,max]
  clustDist = 0
  if(nrow(clustSum) > 1){
    for(i in 2:nrow(clustSum)){
      clustDist = c(clustDist, clustSum[i,min] -clustEnd)
      clustEnd = clustSum[i,max]
    }
  } 
  
  clustSum[,dist := clustDist]
  
  ## repel points
  cdf3 = c()
  clusters = clustSum[,cluster]
  
  for(i in 1:length(clusters)){
    tempClust = cdf2[cluster == clusters[i]]
    #print(tempClust)
    if(i == 1){
      cdf3 = rbind(cdf3, shiftPoints(clust = tempClust, leftDist = tempClust[1,pos], rightDist = clustSum[2, dist]))
      #In case repelled points cross protein length, bring it down to protein len.
      cdf3$pos2 = ifelse(test = cdf3$pos2 >= protLen, yes = protLen, no = cdf3$pos2)
    } else if(i == length(clusters)){
      cdf3 = rbind(cdf3, shiftPoints(clust = tempClust, leftDist = clustSum[i, dist], rightDist = protLen))
      cdf3$pos2 = ifelse(test = cdf3$pos2 >= protLen, yes = protLen, no = cdf3$pos2)
    } else{
      cdf3 = rbind(cdf3, shiftPoints(clust = tempClust, leftDist = clustSum[i, dist], rightDist = clustSum[i+1, dist]))
      cdf3$pos2 = ifelse(test = cdf3$pos2 >= protLen, yes = protLen, no = cdf3$pos2)
    }
  }
  
  return(cdf3)
}