R/genepop_structure.R

In rystanley/genepopedit: Simple and flexible manipulation of genomic data

# Genepop -> STRUCTURE
#' @title Convert Genepop to STRUCTURE format.
#' @description Function to convert Genepop to STRUCTURE
#' @param genepop the genepop data to be manipulated. This can be either a file path
#' or a dataframe read in with tab separation, header=FALSE , quote="", and stringsAsFactors=FALSE.
#' This will be the standard genepop format with the first n+1 rows corresponding to the n loci names,
#' or a single comma delimited row of loci names followed by the locus data. Populations are
#' separated by "Pop". Each individual ID is linked to the locus data by " ,  " (space,space space) and is read in as
#' as a single row (character).
#' @param popgroup if specified (Default: NULL) popgroup is a dataframe or path to a csv.
#' This dataframe contains two columns. Column 1 corresponds to the population names. These names
#' should match the individual IDs (e.g. BON_01 ,  110110 would be 'BON'). The next column
#' has the group. If groupings are the same as populations then leave as NULL (Default).If the input genepop file does not have population and sample ID seperation using ("_") then refer to genepop_ID().
#' @param locusnames logical (default=FALSE). Specify TRUE if you want the locus names from your Genepop file to be the first row in your Structure files.
#' @param path the filepath and filename of output.
#' @rdname genepop_structure
#' @importFrom data.table fread
#' @importFrom utils write.table
#' @importFrom utils read.csv
#' @export


genepop_structure <- function(genepop,popgroup=NULL,locusnames=FALSE,path=NULL){

  #Check to see if genepop is a data.frame from the workspace and convert to data.table
  if(is.data.frame(genepop)){genepop <- as.data.table(genepop)}

  #Check to see if genepop is a file path or dataframe
  if(is.character(genepop)){
    genepop <- data.table::fread(genepop,
                                 header = FALSE, sep = "\t",
                                 stringsAsFactors = FALSE)
  }

  ## check if loci names are read in as one large character vector (1 row)
  header <- genepop[1,]
  if(length(gregexpr(',', header, fixed=F)[[1]])>1){
    lociheader <- strsplit(header,",")
    lociheader <- gsub(" ","",unlist(lociheader))
    #remove the first column of loci names
    genepop <- as.vector(genepop)
    genepop <- genepop[-1,]
    genepop <- c(lociheader,genepop)
    genepop <- as.data.table(genepop,stringsAsFactors = FALSE)
  }

  ## Stacks version information
  stacks.version <- genepop[1,] #this could be blank or any other source. First row is ignored by genepop

  #Remove first label of the stacks version
  genepop <- genepop[-1,]
  colnames(genepop) <- "data"

  #ID the rows which flag the Populations
  Pops  <-  which(genepop$data == "Pop" | genepop$data =="pop" | genepop$data == "POP")
  npops  <-  1:length(Pops)

  ## separate the data into the column headers and the rest
  ColumnData <- genepop$data[1:(Pops[1]-1)]
  ColumnData <- gsub("\r","",ColumnData)#remove any hidden carriage returns
  snpData <- genepop[Pops[1]:NROW(genepop),]

  #Get a datafile with just the snp data no pops
  tempPops <- which(snpData$data=="Pop"| snpData$data =="pop" | snpData$data == "POP") ## Changed because we allowed
  snpData <- snpData[-tempPops,]

  #separate the snpdata
  temp <- as.data.frame(do.call(rbind, strsplit(snpData$data," ")))

  #data format check
  if(unique(temp[,2])!="," | !length(which(temp[,3]==""))>1){
    stop("Genepop sampleID delimiter not in proper format. Ensure sampleIDs are separated from loci by ' ,  ' (space comma space space). Function stopped.",call. = FALSE)
  }
  temp2 <- temp[,4:length(temp)] #split characters by spaces

  #Contingency to see if R read in the top line as the "stacks version"
  if (length(temp2)!=length(ColumnData)){colnames(temp2) <- c(stacks.version,ColumnData)}
  if (length(temp2)==length(ColumnData)){colnames(temp2) <- ColumnData}
  if (length(temp2)!=length(ColumnData)){stacks.version="No STACKS version specified"}

  #stacks version character
  stacks.version <- as.character(stacks.version)

  ## Get the population names (prior to the _ in the Sample ID)
  NamePops <- temp[,1] # Sample names of each
  NameExtract <- substr(NamePops,1,regexpr("_",NamePops)-1)

  #convert the snp data into character format to get rid of factor levels
  temp2[] <- lapply(temp2, as.character)

  #allele coding length
  alleleEx <- max(sapply(temp2[,1],FUN=function(x){nchar(as.character(x[!is.na(x)]))})) #presumed allele length

  #check to make sure the allele length is a even number
  if(!alleleEx %% 2 ==0){stop(paste("The length of each allele is assumed to be equal (e.g. loci - 001001 with 001 for each allele), but a max loci length of", alleleEx, "was detected. Please check data."))}

  #get the allele values summary header
  firstAllele <-  as.data.frame(sapply(temp2,function(x)as.numeric(as.character(substring(x,1,alleleEx/2)))))
  secondAllele <-  as.data.frame(sapply(temp2,function(x)as.numeric(as.character(substring(x,(alleleEx/2)+1,alleleEx)))))

  # switch from combined allele in one row to two allele each in their own row according to a given locus
  holdframe <- rbind(firstAllele,secondAllele)#create a dummy data frame
  holdframe[!1:nrow(holdframe) %% 2 == 0,] = firstAllele #odd rows
  holdframe[1:nrow(holdframe) %% 2 == 0,] = secondAllele #even rows

  holdframe[holdframe==0]= -9 # replace missing values with -9


  # Get the population groupings
  if(!is.null(popgroup)) #if popgroup isn't NULL
  {
    if(is.character(popgroup)){popgroup <- utils::read.csv(popgroup,header=T)} #if it is a path then read it in

    if(length(intersect(unique(NameExtract),popgroup[,1]))!=length(unique(NameExtract))){
      message("Popuation levels missing form popgroups input. STRUCTURE groups now set to default population levels")
      groupvec <- NameExtract
      for (i in 1:length(unique(NameExtract))) # replace with numbers
      {
        groupvec[which(groupvec==unique(NameExtract)[i])] = i
      }
    }

    groupvec=NameExtract
    for (i in 1:nrow(popgroup))
    {
      groupvec[which(groupvec==popgroup[i,1])]=rep(popgroup[i,2],length(groupvec[which(groupvec==popgroup[i,1])]))
    }

  }

  if(is.null(popgroup)) #if popgroup isn't NULL
  {
    groupvec <- NameExtract
    for (i in 1:length(unique(NameExtract))) # replace with numbers
    {
      groupvec[which(groupvec==unique(NameExtract)[i])] = i
    }

  }


  #combine data into single text files and structure format
  holdframe=cbind(rep(NamePops,each=2),rep(groupvec,each=2),holdframe)
  Loci <- do.call(paste,c(holdframe[,], sep=" "))
  headinfo <- paste0("  ",do.call(paste,c(as.list(colnames(temp2)))))
  Output <- data.frame(c(headinfo,Loci))

  if(locusnames==TRUE){
    Output=Output
  } else {
    Output<-Output[-1,]
  }

  #Save Output
  utils::write.table(Output,path,col.names=FALSE,row.names=FALSE,quote=FALSE)

}