ABSTRACT
Recently, Langdon et al. [10] reported that all human Affymetrix microarrays available in the Gene Expression Omnibus (GEO) [11] contain spatial defects to some degree. Thus, quality control for microarrays remains a major issue. Although many methods and software tools have been developed for quality assessment of microarrays [12, 13, 14, 15], detection of spatial artifacts is not yet routinely applied. Furthermore, it is usually not clear how to proceed once such artifacts have been detected. The two alternatives are (1) to either completely exclude or (2) to include the corresponding arrays for any subsequent analysis. In the first case, the corresponding measurements are not available for gene expression profiling and may even have to be repeated if they are crucial to the analysis. This can be cost-intensive, in particular, if corresponding samples have been used up. In the second case, one has to assume that normalization and summarization methods can correct for the measurement errors.
