R/loadData.R

In kevincjnixon/BinfTools: Useful functions for standard analyses

#' Load HTSeq-count data from a sampleTable into a countMatrix and colData object
#'
#' Function to use a sampleTable (same format for DESeqDataSetFromHTSeqCount) to generate a count matrix and colData object
#'
#' @param sampleTable Character indicating the path to a tab-delimited (.txt) or comma-separated (.csv) sampleTable, or a data frame.
#' @return A list of length 2. Counts = count matrix and colData = metadata for samples
#' @export

loadData<-function(sampleTable){
  if(is.character(sampleTable)){
    if(length(grep(".txt", sampleTable))>0 | length(grep(".tsv", sampleTable))>0){
      sampleTable<-read.delim(sampleTable)
    }
    if(length(grep(".csv", sampleTable))>0){
      sampleTable<-read.csv(sampleTable)
    }
  } else {
    if(!is.data.frame(sampleTable)){
      stop("sampleTable should be character or data.frame.")
    }
  }
  x<-list()
  for(i in 1:nrow(sampleTable)){
    x[[i]]<-read.delim(sampleTable[i,2], header=FALSE, row.names=1)
    colnames(x[[i]])<-sampleTable[i,1]
    names(x)[i]<-sampleTable[i,1]
  }
  x<-do.call("cbind", x)
  colData<-sampleTable[,c(1,3:ncol(sampleTable))]
  return(list(Counts=x, colData=colData))
}

#' Function to load featureCounts data
#'
#' @param filename Character indicating the path to the featureCounts table to import
#' @return A list of length 2: Counts=featureCount count matrix and annotation= Gene Annotation information from featureCounts
#' @export

loadFC<-function(filename){
  x<-read.delim(filename, skip=1)
  anno<-x[,c(1:6)]
  counts<-x[,c(7:ncol(x)),drop=FALSE]
  rownames(counts)<-x$Geneid
  colnames(anno)[1]<-"GeneId"
  return(list(Counts=counts, annotation=anno))
}


#Take a numeric vector (i.e. a vector of gene expression) and assign it into named quantiles of a desired length - great for making a 'condition' vector based off of gene expression
#' Generate and assign quantiles
#' Function great for making a 'condition' vector based on gene expression values
#' @param x numeric vector to be used to generate quantiles
#' @param groups character vector of names to be assigned to quantiles in order of ascending values. Number of quantiles generated is based off of length(groups).
#' @return character vector of length(x) with names of groups assigned to each value of x
#' @export

quart_group<-function(x, groups=c("1st","2nd","3rd","4th")){
  q<-quantile(x, probs=seq(0,1, (1/length(groups))))
  g<-c(rep(groups[length(groups)], length(x)))
  for(i in 1:length(g)){
    for(k in (length(q)-1):2){
      if(x[i]<q[k]){
        g[i]<-groups[k-1]
      }
    }
  }
  return(g)
}

#' Combine two count tables
#'
#' @param x data.frame object representing the first count table. Entries must be numeric with columns representing samples and rownames are unique gene names.
#' @param y data.frame object representing the second count table following the same rules as x. Both x and y must have shared rownames (gene) meaning the nomenclature must be the same and there must be at least SOME shared genes. Genes not shared between samples will be removed. Gene order does not matter.
#' @param xName Character indicating an optional identifier added to the colnames of x. Default is NULL.
#' @param yName Character indicating an optional identifier added to the colnames of y. Default is NULL.
#' @return data.frame object of combined x and y where the rownames are shared genes and columns are samples.
#' @export

combCounts<-function(x,y, xName=NULL, yName=NULL){
  genes<-rownames(x)[which(rownames(x) %in% rownames(y))]
  genes<-rownames(y)[which(rownames(y) %in% genes)]
  message(length(genes)," genes shared")
  x<-x[which(rownames(x) %in% genes),]
  y<-y[which(rownames(y) %in% genes),]
  if(!is.null(xName)){
    colnames(x)<-paste(colnames(x), xName, sep=".")
  }
  if(!is.null(yName)){
    colnames(y)<-paste(colnames(y), yName, sep=".")
  }
  return(cbind(x, y[match(rownames(x), rownames(y)),]))
}