This is a collection of methods for (R)IP-chip (E)nrichment (A)nalysis. A table of normalized expression values from a RIP-chip experiment is read (e.g. as produced by the Affymetrix PowerTools) and two main steps are performed: Subtraction of the background of non target genes and secondary normalization for differing IP efficiencies.
The background of non-target genes is identified by applying gaussian mixture modelling (see Figure). In our paper, we show that this is not only appropriate but a necessary step in the analysis of such kind of data:
Background subtracted enrichment values are closer to a reference standard.
Computed false discovery rates are valid based on the reference.
It allows to identify non-expressed genes.
The secondary normalization is done using PCA and in our paper we show that this normalization results in better performance for both target identification and differential target identification.