You want to talk about single genes and not lists of genes

Now you have a list of genes and you know that about 10% of them are false positives. You want to know which ones. This is of course not possible. On the other hand, the gene on top of the list is less likely to be a false positive than the one on the cutoff line. Efron et al. (7) introduced the local FDR (LFDR) which is the probability that a gene is a false positive. Note, that the p-value is a different probability. For computing the LFDR, apply function twilight which is based on the estimation procedure described in Scheid and Spang (15).

LFDR <- twilight(score)

Plot the LFDR over the range of p-values by calling plot(LFDR, "fdr"). Following the LFDR from low p-values to high p-values, you get an impression of the level of differential gene expression in your experiment, and whether there is a twilight zone where clear differential expression fades into clear non-differential expression.

0 0

Post a comment