When running BioQC on an expression set, take only rows with gene symbol as background.

[grst]

because of the reasons discussed...

I'm not sure what should be the default behavior, though. But taking all genes as background could result in misleading results and significant 'over-estimation' of signatures.

[grst]

This function can be used to filter expression sets accordingly:

#' Filter genes with annotation from expression set. 
#' 
#' This is done to have the correct background for BioQC.
#' Taking all probeids as background is a bias towards each signature in general 
#' as probeids with gene symbol tend to be higher expressed in general. 
filter_eset = function(eset) {
  gene_symbols = fData(eset)$BioqcGeneSymbol
  # remove lines that have no gene set
  eset = eset[(!is.na(gene_symbols)) & gene_symbols != '-',]
  eset = eset[keepMaxStatRowInd(exprs(eset), fData(eset)$BioqcGeneSymbol),]
  return(eset)
}

[grst]

Also, we need to check that the gene symbols are valid hgnc symbols.

My current version of this function looks like this:

#' Filter genes with annotation from expression set. 
#' 
#' This is done to have the correct background for BioQC.
#' Taking all probeids as background is a bias towards each signature in general 
#' as probeids with gene symbol tend to be higher expressed in general. 
filter_eset = function(eset) {
  gene_symbols = fData(eset)$BioqcGeneSymbol
  hgnc_symbols = read_tsv("results/hgnc_symbols.tsv")
  # remove lines that have no gene set
  eset = eset[(!is.na(gene_symbols)) & (gene_symbols != '-') & (gene_symbols %in% hgnc_symbols$hgnc_symbols),]
  eset = eset[keepMaxStatRowInd(exprs(eset), fData(eset)$BioqcGeneSymbol),]
  return(eset)
}

where hgnc_symbols.tsv is derived from ftp://ftp.ebi.ac.uk/pub/databases/genenames/new/tsv/hgnc_complete_set.txt