R/RNASeqAssay-methods.R

In MAST: Model-based Analysis of Single Cell Transcriptomics

## Code from Masanao.  Not sure about what it is supposed to do.
## ##' Filter an RNASeqAssay
## ##'
## ##' Use the percentage of library per gene CDF type plots
## ##' to pick out outliers using the \code{1.5*IQR}
## ##' threshold
## ##' @param sca a \code{SingleCellAssay}
## ##' @param plot a \code{logical} indicating whether plots should be generated
## ##' @export
## ##' @return a filtered SingleCellAssay
## outlierDetection <- function(sca,plot=TRUE){
##     if(!is(sca,"SingleCellAssay")){
##         stop(sprintf("sca must be an SingleCellAssay, but found %s",class(sca)[1]))
##     }
    
##                                         #compute the %of total library size for each gene
##     ee <- assay(sca)
##     f <- t(t(ee) / (colSums(ee)))
    
##                                         #filter out unexpressed genes
##     gfilt <- apply(f,2,sum, na.rm=TRUE)!=0
##     f <- f[,gfilt]
    
##                                         #cumulative sum by library
##     fc <- t(apply(f,1,cumsum))
    
##                                         #median
##     md <- apply(fc,1,median)
    
##                                         #filter at 1.5*IQR
##     filter <- (md>(median(md)+1.5*IQR(md))|md<(median(md)-1.5*IQR(md)))
    
##     if(plot){
        
##         #'plot the outliers
##         par(mfrow=c(1,2))
##         plot(y=md,x=rank(md),main="Outliers",xlab="rank",ylab="median frequency")
##         points(y=md[filter],x=rank(md)[filter],col="red")
        
##                                         #cdfs
##         plot(fc[1,],type="l",main="Outliers",xlab="rank",ylab="cdf")
##         apply(fc,1,lines)
##         apply(fc[filter,],1,function(x)lines(x,col="red"))
##     }
##                                         #return the filtered object
##     return(sca[!filter,gfilt])
## }


#'Filter low-expressing genes
#'
#'Filter out genes that have less than some percent threshold expression across all libraries
#'
#'@param assay a \code{SingleCellAssay} object
#'@param threshold a \code{numeric} between 0, and 1, specifying the threshold frequency below which genes will be filtered out
#' @return \code{SingleCellAssay}
#'@export
#' @examples
#' data(vbetaFA)
#' filterLowExpressedGenes(vbetaFA)
filterLowExpressedGenes<-function(assay,threshold=0.1){
    if(!is(assay,"SingleCellAssay")){
        stop(sprintf("assay must be a SingleCellAssay but found %s\n",class(assay)[1]))
    }
    gp <- freq(assay)
    cat(sprintf("Filtered out %2.2f percent of genes at %2.2f percent threshold\n",mean(gp>threshold)*100,threshold*100))
    return(assay[gp>threshold,])
}