pvalueFilter-method: Returns a dataframe with results of the analysis for a...
In vodkatad/roar: Identify differential APA usage from RNA-seq alignments

Description Usage Arguments Value Examples

The last step of a classical Roar analyses: it returns a dataframe containing m/M values, roar values, pvalues and estimates of expression (a measure recalling FPKM). Only the genes with an expression estimate bigger than a given cutoff will be considered. Also pvalues will be considered for filtering.

1 2	pvalueFilter(rds, fpkmCutoff, pvalCutoff)

`rds`	The `RoarDataset` or the `RoarDatasetMultipleAPA` with all the analysis steps (`countPrePost`, `computeRoars`, `computePvals`) performed. If one or more steps hadn't been performed they will be called automatically.
`fpkmCutoff`	The cutoff that will be used to determine if a gene is expressed or not.
`pvalCutoff`	The cutoff that will be used to determine if a pvalue is significative or not.

For RoarDataset:

The resulting dataframe will be identical to that returned by standardFilter but after gene expression and m/M values filtering another step will be performed: for single samples comparisons only genes with a nominal pvalue smaller than the given cutoff will be considered, while for multiple samples a column (nUnderCutoff) will be added to the dataframe. This column will contain an integer number representing the number of comparisons between the samples of the two conditions that results in a nominal pvalue lower than the given cutoff (pvalCutoff). For multiple samples with a paired design (i.e. if computePairedPvals was used) the pvalues of the requested pairings will be listed together with the combined pvalued obtained with the Fisher method and the filtering will be done on this pvalue.

For RoarDatasetMultipleAPA: for each gene we select the APA choice that is associated with the smallest p-value then proceed exactly as above.

   library("GenomicAlignments")
   gene_id <- c("A_PRE", "A_POST", "B_PRE", "B_POST")
   features <- GRanges(
      seqnames = Rle(c("chr1", "chr1", "chr2", "chr2")),
      strand = strand(rep("+", length(gene_id))),
      ranges = IRanges(
         start=c(1000, 2000, 3000, 3600),
         width=c(1000, 900, 600, 300)),
      DataFrame(gene_id)
   )
   rd1 <- GAlignments("a", seqnames = Rle("chr1"), pos = as.integer(1000), cigar = "300M", strand = strand("+"))
   rd2 <- GAlignments("a", seqnames = Rle("chr1"), pos = as.integer(2000), cigar = "300M", strand = strand("+"))
   rd3 <- GAlignments("a", seqnames = Rle("chr2"), pos = as.integer(3000), cigar = "300M", strand = strand("+"))
   rds <- RoarDataset(list(c(rd1,rd2)), list(rd3), features)
   rds <- countPrePost(rds, FALSE)
   rds <- computeRoars(rds)
   rds <- computePvals(rds)
   dat <- pvalueFilter(rds, 1, 0.05)