Outlier resistance, standardization, and modeling issues for DNA microarray data

Gene expression levels for several viral genes placed on DNA microarrays were measured under mock and true infection with the objective of identifying which viral genes are upregulated by infection, i.e., are expressed to a higher degree when exposed to true infection as compared to noninfection. After nonlinearly standardizing the microarrays to a single intensity scale, the average expression level for each gene was estimated resistantly using a double median estimator. Approaches based on mixed effects modeling allows strength to be borrowed across genes to assess the significance of any evidence of upregulation.