This paper presents a statistical model and method for background adjustment in high-density oligonucleotide expression arrays, specifically Affymetrix GeneChip arrays. The authors propose a formal statistical framework to improve the default background adjustment algorithms provided by Affymetrix. The model uses probe sequence information to account for non-specific hybridization and optical noise, which are significant sources of error in gene expression measurements. The proposed method, denoted as GC-RMA, is compared with other popular methods such as MAS 5.0 and RMA using both simulated and real data. The results show that GC-RMA provides better accuracy and precision, especially for genes with low expression levels. The software for implementing the background adjustment is available as part of the Bioconductor project. The authors also discuss potential future improvements and limitations of their approach.This paper presents a statistical model and method for background adjustment in high-density oligonucleotide expression arrays, specifically Affymetrix GeneChip arrays. The authors propose a formal statistical framework to improve the default background adjustment algorithms provided by Affymetrix. The model uses probe sequence information to account for non-specific hybridization and optical noise, which are significant sources of error in gene expression measurements. The proposed method, denoted as GC-RMA, is compared with other popular methods such as MAS 5.0 and RMA using both simulated and real data. The results show that GC-RMA provides better accuracy and precision, especially for genes with low expression levels. The software for implementing the background adjustment is available as part of the Bioconductor project. The authors also discuss potential future improvements and limitations of their approach.