There are many references about the microarray data normalization, the limma package utilized most of the available normalization methods and NTAP didn't provide any additional algorithm for the normalization procedure. However, please be aware that some assumptions for the normalization methods are not true anymore in the case of tiling array. For example, for the expression array to detect differential expressed genes between different tissues, growth conditions or treatments, we can assume that the majority of the genes in the whole genome whose expression level wouldn't change in the experiment, and the number of increased and decreased genes are comparable. In the tiling array applications like ChIP-chip data, the assumption could be wrong. First of all, there are definitely more increased signal because we enriched the TF-binding regions or histone modification regions in the sample preparation step. Furthermore, the portion of oligos who represent the modification or binding are not known, it could be quite large in one case, but very few for another.

We recommend to use the most conservative normalization method and always check the plots before and after the normalization to see whether the normalization effect is good or bad.