Background correction using dinucleotide affinities improves the performance of GCRMA

Area of Science:

• Bioinformatics
• Genomics
• Molecular Biology

Background:

• High-density short oligonucleotide microarrays are essential for global gene expression analysis.
• Background noise in microarray data can significantly impact data interpretation.
• Accurate estimation of background noise is crucial for reliable gene expression studies.

Purpose of the Study:

• To develop an improved method for calculating probe affinity on microarrays.
• To incorporate nearest-neighbor (NN) dinucleotide information into probe affinity models.
• To enhance the accuracy of gene expression analysis by correcting for background noise.

Main Methods:

• Developed a novel approach to calculate probe affinity based on sequence composition and NN information.
• Utilized position-specific dinucleotide data, surpassing previous single nucleotide models.
• Integrated the new affinity model into the GCRMA (GeneChip Robust Multi-array Average) preprocessing algorithm.

Main Results:

• The new model explained up to 10% more variance (R2) compared to existing models.
• Correcting for background noise using the dinucleotide affinity model improved GCRMA performance on control datasets.
• Enhanced detection of low-intensity targets was observed with the improved background correction.

Conclusions:

• Incorporating dinucleotide information into position-dependent affinity models significantly enhances model performance.
• The dinucleotide affinity model improves the detection of differentially expressed genes when used for background correction in GeneChip preprocessing.
• The findings align with physical models of binding affinity, emphasizing the role of nearest-neighbor stacking interactions.