I am trying to normalize this dataset (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE88884) which uses the HTA 2.0 platform with a custom CDF from BrainArray. I then plan to train a k-Top Scoring Pairs classifier for use on another platform.

Which preprocessing method will best reflect the rank order of gene expression within each array?

I have tried SCAN as it would account for probe composition and background noise; however, it is quite slow for this platform and number of samples (15 min/sample).