Limit of detection analysis for ExpressionSet objects

Computes the percentage of values below a feature-specific lower limit of detection (LOD) in an ExpressionSet. Commonly used in proteomic quality assessment.

Usage

get.prop.below.LLOD(eset, flagged = c("OUT", "IN"))

Arguments

eset

An ExpressionSet object containing expression values and feature metadata. Must include a numeric column lod.max in fData(eset).

flagged

Character string indicating whether flagged samples should be removed. One of "OUT" (remove flagged samples) or "IN" (retain all samples).

Value

An ExpressionSet with an added feature annotation column:

pc.belowLOD.new

Percentage of values below the LOD per feature.

Details

The function compares expression values from exprs(eset) to feature-specific LOD thresholds stored in fData(eset)$lod.max. Computation is vectorised using sweep() for efficiency and consistency with Bioconductor conventions.

Flagged samples (if present in pData(eset)$Flagged) can optionally be removed.

Examples

if (FALSE) { # \dontrun{
suppressMessages(library(Biobase))
data(sample.ExpressionSet, package = "Biobase")
exampleSet <- sample.ExpressionSet
set.seed(1)
probs <- runif(nrow(exampleSet))
Biobase::fData(exampleSet)$lod.max <- vapply(seq_len(nrow(exampleSet)), function(i)
           quantile(Biobase::exprs(exampleSet)[i, ], probs = probs[i], na.rm = TRUE),
           numeric(1))
lod <- get.prop.below.LLOD(exampleSet)
fd <- Biobase::fData(lod)
fd <- fd[order(fd$pc.belowLOD.new, decreasing = TRUE), ]
knitr::kable(head(fd))
plot(fd$pc.belowLOD.new, main = "Random quantile cut off", ylab = "% below LOD")
} # }