Monday, 18 October 2010

NUSE and RLE Plots

These are diagnostic plots that can be created from the affyPLM package.

Nuse is the Normalised Unscaled Standard Error and RLE is the Relative Log Expression. These comparisons only work within datasets - the same arrays and the experimental protocols and not between datasets. Arrays are outliers if they have very different NUSE values to all of the others or if they have a wider distribution.

dataPLM <- fitPLM(raw)
boxplot(dataPLM, main="NUSE",ylim = c(0.95, 1.10),
  outline = FALSE, col="lightblue",las=3,
  whisklty=0, staplelty=0)

Mbox(dataPLM, main="RLE", ylim = c(-0.4, 0.4),
  outline = FALSE, col="mistyrose", las=3,
  whisklty=0, staplelty=0)

In my particular case there is one array that has poor values for RLE and NUSE and that is array GSM372776.CEL. This can be removed from subsequent analysis by creating a new object without this array.

badArray = match("GSM372776.CEL", sampleNames(raw))
raw1 <- raw[,-badArray]

No comments:

Post a Comment