filterDEGs: Filter and plot DEG results

Description

Filters and plots DEG results for a given set of sample comparisons. The gene idenifiers of all (i) Up_or_Down, (ii) Up and (iii) Down regulated genes are stored as separate list components, while the corresponding summary statistics, stored in a fourth list component, is plotted in form of a stacked bar plot.

Usage

filterDEGs(degDF, filter, plot = TRUE)

Arguments

degDF

data.frame generated by run_edgeR

filter

Named vector with filter cutoffs of format c(Fold=2, FDR=1) where Fold refers to the fold change cutoff (unlogged) and FDR to the p-value cutoff.

plot

Allows to turn plotting behavior on and off with default set to TRUE.

Value

Returns list with four components
UporDownList of up or down regulated gene/transcript indentifiers meeting the chosen filter settings for all comparisons defined in data frames pval and log2FC.
UpSame as above but only for up regulated genes/transcript.
DownSame as above but only for down regulated genes/transcript.

Details

Currently, there is no community standard available how to calculate fold changes (here logFC) of genomic ranges, such as gene or feature ranges, to unambiguously refer to them as features with increased or decreased read abundandce; or in case of gene expression experiments to up or down regulated genes, respectively. To be consistent within systemPipeR, the corresponding functions, such as filterDEGs, use here the following definition. Genomic ranges with positive logFC values are classified as up and those with negative logFC values as down. This means if a comparison among two samples a and b is specified in the corresponding targets file as a-b then the feature with a positive logFC has a higher _normalized_ read count value in sample a than in sample b, and vice versa. To inverse this assignment, users want to change the specification of their chosen sample comparison(s) in the targets file accordingly, e.g. change a-b to b-a. Alternatively, one can swap the column order of the matrix assigned to the cmp argument of the run_edgeR or run_DESeq2 functions. Users should also be aware that for logFC values close to zero (noise range), the direction of the fold change (sign of logFC) can be very sensitive to minor differences in the normalization method, while this assignment is much more robust for more pronounced changes or higher absolute logFC values.

Examples

Run this code

targetspath <- system.file("extdata", "targets.txt", package="systemPipeR")
targets <- read.delim(targetspath, comment="#")
cmp <- readComp(file=targetspath, format="matrix", delim="-")
countfile <- system.file("extdata", "countDFeByg.xls", package="systemPipeR")
countDF <- read.delim(countfile, row.names=1)
edgeDF <- run_edgeR(countDF=countDF, targets=targets, cmp=cmp[[1]], independent=FALSE, mdsplot="")
pval <- edgeDF[, grep("_FDR$", colnames(edgeDF)), drop=FALSE]
fold <- edgeDF[, grep("_logFC$", colnames(edgeDF)), drop=FALSE]
DEG_list <- filterDEGs(degDF=edgeDF, filter=c(Fold=2, FDR=10))
names(DEG_list)
DEG_list$Summary