Oligonucleotide arrays: information from replication and spatial structure.

The introduction of oligonucleotide DNA arrays has resulted in much debate concerning appropriate models for the measurement of gene expression. By contrast, little account has been taken of the possibility of identifying the physical imperfections in the raw data.This paper demonstrates that, with the use of replicates and an awareness of the spatial structure, deficiencies in the data can be identified, the possibility of their correction can be ascertained and correction can be effected (by use of local scaling) where possible. The procedures were motivated by data from replicates of Arabidopsis thaliana using the GeneChip ATH1-121501 microarray. Similar problems are illustrated for GeneChip Human Genome U133 arrays and for the newer and larger GeneChip Wheat Genome microarray.R code is freely available on request.