R/convertToRangedSummarizedExperiment.R

In intermine/intermineR: R Interface with InterMine-Powered Databases

#' @import IRanges
#' @import GenomicRanges
#' @import SummarizedExperiment
#' @export

# Define function for converting InterMine experimental data 
# to RangedSummarizedExperiment
# Dataset argument takes data.frame object as input

convertToRangedSummarizedExperiment = function(
  im,
  dataset,
  SampleColumn,
  GeneColumn,
  ValueColumn,
  OrganismValue,
  colsForSampleMetadata,
  exonsForRowRanges = FALSE
){
  
  # get index for SampleColumn and GeneColumn
  #if(class(SampleColumn) %in% c("integer", "numeric")){
  if(is.integer(SampleColumn) | is.numeric(SampleColumn)){
    sc = SampleColumn
  #} else if (class(SampleColumn) == "character"){
  } else if (is.character(SampleColumn)){
    sc = which(colnames(dataset) == SampleColumn)
  }
  
  #if(class(GeneColumn) %in% c("integer", "numeric")){
  if(is.integer(GeneColumn) | is.numeric(GeneColumn)){
    gc = GeneColumn
  #} else if (class(GeneColumn) == "character"){
  } else if (is.character(GeneColumn)){
    gc = which(colnames(dataset) == GeneColumn)
  }
  
  #if(class(ValueColumn) %in% c("integer", "numeric")){
  if(is.integer(ValueColumn) | is.numeric(ValueColumn)){
    vc = ValueColumn
  #} else if (class(ValueColumn) == "character"){
  } else if (is.character(ValueColumn)){
    vc = which(colnames(dataset) == ValueColumn)
  }
  
  # make sample names unique for each gene
  unique.sample.names = c()
  for(j in unique(dataset[,gc])){
    
    # index for samples per gene
    ind.gene = dataset[,gc] == j
    
    # assign suffixes to make sample names unique
    unique.sample.names = c(unique.sample.names,
                            make.unique(dataset[ind.gene,sc])
    )
  }
  
  dataset[,sc] = unique.sample.names
  
  # get count.table for RangedSummarizedExperiment object with 
  # the reshape function
  count.table = reshape(
    data = dataset[,c(gc, sc, vc)],
    idvar = colnames(dataset)[sc],
    timevar = colnames(dataset)[gc],
    direction = "wide"
  )
  
  # set sample names in rows
  rownames(count.table) = count.table[,1]
  count.table = count.table[,2:ncol(count.table)]
  
  # set gene names in columns
  colnames(count.table) = unique(dataset[,gc])
  
  # convert count.table NA values to zero and all values to class numeric
  for(i in seq(ncol(count.table))){
    ind = is.na(count.table[,i])
    count.table[ind,i] = "0"
    count.table[,i] = as.numeric(count.table[,i])
  }
  
  # reverse count.table to set samples in columns and genes in rows!
  count.table = t(count.table)
  
  # create GRanges object for the rowRanges of RangedSummarizedExperiment object
  # exonsForRowRanges is used to define whether exon chromosome locations 
  # will be used for rowRanges argument instead of gene chromosome locations 
  # (default)
  # if exonsForRowRanges = TRUE then:
  # create new InterMineR query to retrieve the exons of each gene
  # the results will be converted to GRanges and subsequently to GRangesList 
  # object for rowRanges
  
  # define new query
  geneExonQuery = newQuery()
  
  # check for correct child names in mine data model
  #model = getModel(im)
  
  strand = "strand"
  start = "start"
  end = "end"
  name = "name"
  symbol = "symbol"
  secondaryIdentifier = "secondaryIdentifier"
  primaryIdentifier = "primaryIdentifier"
  
  #if(any(model$child_name == "strand")){
  #  strand = "strand"
  #} else if(any(model$child_name == "Strand")){
  #  strand = "Strand"
  #}
  #
  #if(any(model$child_name == "start")){
  #  start = "start"
  #} else if(any(model$child_name == "Start")){
  #  start = "Start"
  #}
  #
  #if(any(model$child_name == "end")){
  #  end = "end"
  #} else if(any(model$child_name == "End")){
  #  end = "End"
  #}
  #
  #if(any(model$child_name == "name")){
  #  name = "name"
  #} else if(any(model$child_name == "Name")){
  #  name = "Name"
  #}
  #
  #if(any(model$child_name == "symbol")){
  #  symbol = "symbol"
  #} else if(any(model$child_name == "Symbol")){
  #  symbol = "Symbol"
  #}
  #
  #if(any(model$child_name == "secondaryIdentifier")){
  #  secondaryIdentifier = "secondaryIdentifier"
  #} else if(any(model$child_name == "Secondary Identifier")){
  #  secondaryIdentifier = "Secondary Identifier"
  #}
  #
  #if(any(model$child_name == "primaryIdentifier")){
  #  primaryIdentifier = "primaryIdentifier"
  #} else if(any(model$child_name == "Primary Identifier")){
  #  primaryIdentifier = "Primary Identifier"
  #}
  
  # set sort order
  order.vector = "ASC"
  names(order.vector) = paste("Gene",secondaryIdentifier, sep = ".")
  
  geneExonQuery$orderBy = list(
    order.vector
  )
  
  # set organism constraint
  organismConstraint = list(
    path = paste("Gene.organism", name, sep = "."),
    op = "=",
    value = OrganismValue,
    code = "B"
  )
  
  if(!exonsForRowRanges){
    geneExonQuery$select = c(
      paste(sep = ".", "Gene", symbol),
      paste(sep = ".", "Gene", secondaryIdentifier),
      paste(sep = ".", "Gene.chromosomeLocation", start),
      paste(sep = ".", "Gene.chromosomeLocation", end),
      paste(sep = ".", "Gene.chromosome", primaryIdentifier),
      paste(sep = ".", "Gene.locations", strand)
    )
    
    l.exons = list(NULL)
    ind.names = c()
    for(j in seq(length(unique(dataset[,gc])))){
      
      g = unique(dataset[,gc])[j]
      
      # set gene constraint
      geneConstraint = list(
        path = 'Gene',
        op = "LOOKUP",
        value = g,
        code = "A"
      )
      
      # set constraints
      geneExonQuery$where = list(
        geneConstraint,
        organismConstraint
      )
      
      # run query
      d = runQuery(im, geneExonQuery)
      
      # convert exon results to GRanges object
      d2 = convertToGRanges(
        dataset = d,
        seqnames = paste(sep = ".", "Gene.chromosome", primaryIdentifier),
        start = paste(sep = ".","Gene.chromosomeLocation",start),
        end = paste(sep = ".","Gene.chromosomeLocation",end),
        strand = paste(sep = ".", "Gene.locations", strand),
        names = paste(sep = ".", "Gene.chromosome", primaryIdentifier),
        columnsAsMetadata = c(
          paste(sep = ".", "Gene", symbol),
          paste(sep = ".", "Gene", secondaryIdentifier)
        )
      )
      
      l.exons[[j]] = d2
      ind.names = c(ind.names, g)
    }
    
  } else {
    # set columns for the query
    geneExonQuery$select = c(
      paste(sep=".","Gene.exons.chromosomeLocation",start),
      paste(sep=".","Gene.exons.chromosomeLocation",end),
      paste(sep=".","Gene.exons.chromosomeLocation",strand),
      paste(sep=".","Gene.exons.chromosome",primaryIdentifier),
      paste(sep=".","Gene",symbol),
      paste(sep=".","Gene",secondaryIdentifier)
    )
    
    l.exons = list(NULL)
    ind.names = c()
    for(j in seq(length(unique(dataset[,gc])))){
      
      g = unique(dataset[,gc])[j]
      
      # set gene constraint
      geneConstraint = list(
        path = 'Gene',
        op = "LOOKUP",
        value = g,
        code = "A"
      )
      
      # set constraints
      geneExonQuery$where = list(
        geneConstraint,
        organismConstraint
      )
      
      # run query
      d = runQuery(im, geneExonQuery)
      
      # convert exon results to GRanges object
      d2 = convertToGRanges(
        dataset = d,
        seqnames = paste(sep=".","Gene.exons.chromosome",primaryIdentifier),
        start = paste(sep=".","Gene.exons.chromosomeLocation",start),
        end = paste(sep=".","Gene.exons.chromosomeLocation",end),
        strand = paste(sep=".","Gene.exons.chromosomeLocation",strand),
        names = paste(sep=".","Gene.exons.chromosome",primaryIdentifier),
        columnsAsMetadata = c(
          paste(sep = ".", "Gene", symbol),
          paste(sep = ".", "Gene", secondaryIdentifier)
        )
      )
      
      l.exons[[j]] = d2
      ind.names = c(ind.names, g)
    }
    
  }
  
  # convert to GRangesList
  l.exons = GRangesList(l.exons)
  names(l.exons) = ind.names
  
  # create DataFrame object for colData of RangedSummarizedExperiment object
  colMetaData = list(NULL)
  
  for(i in colsForSampleMetadata){
    
    j = which(colsForSampleMetadata == i)
    
    colMetaData[[j]] = tapply(
      X = dataset[,i],
      INDEX = dataset[,sc],
      function(x){
        
        if(length(unique(x)) == 1){
          y = unique(x)
        } else {
          y = paste0(unique(x), collapse = " & ")
        }
        
        return(y)
      }
    )
    
  }
  
  # cbind data to data.frame
  colMetaData = do.call(cbind, colMetaData)
  
  # change column names
  colnames(colMetaData) = colnames(dataset)[colsForSampleMetadata]
  
  # set the rows of colMetaData to the same order as the columns of count.table
  ind.colMetaData = c()
  for(i in seq(ncol(count.table))){
    
    ind.colMetaData = c(ind.colMetaData, 
                        which(rownames(colMetaData) == colnames(count.table)[i])
    )
    
  }
  
  colMetaData = colMetaData[ind.colMetaData,]
  
  # convert to DataFrame class
  colMetaData = DataFrame(colMetaData)
  
  # create RangedSummarizedExperiment object
  result = SummarizedExperiment(
    assays = list(counts = count.table),
    rowRanges = l.exons,
    colData = colMetaData
  )
  
  return(result)
}