Statistical analysis of high-density oligonucleotide arrays: a multiplicative noise model

Area of Science:

• Bioinformatics
• Genomics
• Statistical Modeling

Background:

• High-density oligonucleotide arrays (GeneChips) are vital for monitoring thousands of genes simultaneously.
• Accurate analysis of GeneChip data is crucial for biomedical research.
• Existing methods like Affymetrix Microarray Suite 5.0 (AMS), Li and Wong (LW), and Naef et al. (FN) have limitations in handling different experimental designs and noise models.

Purpose of the Study:

• To develop a novel, model-based method for GeneChip analysis.
• To improve the accuracy and reliability of gene expression quantification.
• To provide a superior analytical tool for experiments involving a series of samples.

Main Methods:

• A statistical model assuming multiplicative noise, unlike additive noise models.
• Elimination of statistically significant outliers to enhance result robustness.
• Estimation of uniform background using both mismatch and perfect match probe intensities, avoiding probe-specific background assumptions.

Main Results:

• The new method demonstrated superior performance compared to AMS and LW.
• It replaced binary 'presence' calls with statistically significant p-values, reducing ambiguity.
• Achieved a 4.4% Type I error rate at the 1.25-fold level, significantly lower than AMS (29%) and LW (15%).
• Showcased a 3.4-times better signal-to-noise ratio than AMS and 1.4-times better than LW.

Conclusions:

• The developed method offers enhanced accuracy and reliability for GeneChip data analysis.
• It provides a more robust alternative for researchers analyzing gene expression in series of samples.
• The method's improved signal-to-noise ratio aids in more precise biological interpretations.